CN115660464A

CN115660464A - Intelligent equipment maintenance method and terminal based on big data and physical ID

Info

Publication number: CN115660464A
Application number: CN202211261182.9A
Authority: CN
Inventors: 王兆丰; 徐智新; 李少鹏; 郭铧; 曲振旭; 谢鹏; 赵鑫成; 郭威; 高翔; 张峻伟; 刘兴宇; 吴一凡
Original assignee: Super High Voltage Branch Of State Grid Fujian Electric Power Co ltd; State Grid Fujian Electric Power Co Ltd
Current assignee: Super High Voltage Branch Of State Grid Fujian Electric Power Co ltd; State Grid Fujian Electric Power Co Ltd
Priority date: 2022-10-14
Filing date: 2022-10-14
Publication date: 2023-01-31

Abstract

The invention discloses an intelligent equipment maintenance method and terminal based on big data and physical ID. In the equipment maintenance process, semantic analysis is carried out on historical defect information of equipment to be maintained, defect treatment opinions are associated with the equipment to be maintained, after the equipment is maintained, a maintenance model is established based on the defect scores and the defect treatment opinions of the equipment to be maintained, and a maintenance plan of the equipment to be maintained is generated according to the maintenance model, so that the positioned equipment can be automatically subjected to defect detection and matched with a corresponding solution, and the equipment maintenance is automatically and intelligently carried out.

Description

Intelligent equipment maintenance method and terminal based on big data and physical ID

Technical Field

The invention relates to the technical field of electric power ultrahigh voltage equipment maintenance, in particular to an intelligent equipment maintenance method and terminal based on big data and physical ID.

Background

The transformer substation is used as an important component of a power grid, the stability of the transformer substation is related to the safe operation of the whole power grid, once a fault occurs, a power grid accident is easily caused, the overhaul of transformer substation equipment is not in place, hidden dangers are not eliminated, and the fault of the transformer substation is often caused. Therefore, the method is very important for timely overhauling the substation equipment.

In the PMS (equipment management terminal) of the current state network company, in the process of a transformer substation maintenance flow, operation and maintenance personnel are required to enter the transformer substation, equipment is checked one by one, then the equipment state is judged according to the experience of the maintenance personnel, and meanwhile, a maintenance report is required to be compiled in the maintenance process, and the PMS terminal is recorded after maintenance is completed.

The maintenance process has various equipment types, hundreds of equipment types and manufacturers, different parameters, and judgment of partial defects and hidden dangers, which usually depends on the technical level of maintenance personnel. In the actual operation and maintenance process, the levels of the maintainers are uneven, so that different maintainers of the same equipment can overhaul the equipment, and different conclusions can be given.

Therefore, an intelligent research and judgment terminal depending on big data is urgently needed, on one hand, the defects are preliminarily and automatically judged according to the artificial intelligence technology, on the other hand, a same type of equipment defect gallery is automatically provided according to the equipment types for the reference of maintainers, and therefore the standard unification in the equipment maintenance process is realized.

Disclosure of Invention

The technical problem to be solved by the invention is as follows: the intelligent equipment maintenance method and terminal based on the big data and the real object ID are provided, and the equipment problem can be quickly positioned and the corresponding solution can be matched.

In order to solve the technical problems, the invention adopts the technical scheme that:

an intelligent equipment maintenance method based on big data and physical ID comprises the following steps:

acquiring equipment to be overhauled obtained by scanning equipment real object ID, and searching corresponding historical defect information according to the information of the equipment to be overhauled;

using a defect metering model for the historical defect information of the equipment to be overhauled, and carrying out defect scoring on the equipment to be overhauled based on data analysis and data statistics;

performing semantic analysis on the historical defect information of the equipment to be overhauled, and associating corresponding defect processing opinions for the equipment to be overhauled;

and establishing a maintenance model based on the defect score and the defect treatment suggestion of the equipment to be maintained, generating a maintenance plan of the equipment to be maintained according to the maintenance model, and using the maintenance plan to maintain the scanned equipment to be maintained.

In order to solve the technical problem, the invention adopts another technical scheme as follows:

an intelligent overhaul terminal of equipment based on big data and physical ID, comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the following steps:

The invention has the beneficial effects that: before the equipment maintenance is carried out, the equipment to be maintained obtained by scanning the equipment physical ID is obtained, corresponding historical defect information is searched according to the information of the equipment to be maintained, and the equipment to be maintained is scored for defects based on a defect metering model, so that the equipment problem can be quickly positioned. In the equipment maintenance process, semantic analysis is carried out on historical defect information of equipment to be maintained, defect treatment opinions are associated with the equipment to be maintained, after the equipment is maintained, a maintenance model is established based on the defect scores and the defect treatment opinions of the equipment to be maintained, and a maintenance plan of the equipment to be maintained is generated according to the maintenance model, so that the positioned equipment can be automatically subjected to defect detection and matched with a corresponding solution, and the equipment maintenance is automatically and intelligently carried out.

Drawings

FIG. 1 is a flowchart of an intelligent overhaul method of a device based on big data and physical ID according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of an intelligent overhaul terminal of a device based on big data and physical ID according to an embodiment of the present invention;

FIG. 3 is a block diagram of an intelligent overhaul method of a device based on big data and physical ID according to an embodiment of the present invention;

description of reference numerals:

1. an intelligent equipment maintenance terminal based on big data and physical ID; 2. a memory; 3. a processor.

Detailed Description

In order to explain technical contents, achieved objects, and effects of the present invention in detail, the following description is made with reference to the accompanying drawings in combination with the embodiments.

Referring to fig. 1, an embodiment of the present invention provides an intelligent overhaul method for a device based on big data and physical ID, including:

performing semantic analysis on the historical defect information of the equipment to be overhauled, and associating corresponding defect processing suggestions for the equipment to be overhauled;

From the above description, the beneficial effects of the present invention are: before the equipment maintenance is carried out, the equipment to be maintained obtained by scanning the equipment physical ID is obtained, corresponding historical defect information is searched according to the information of the equipment to be maintained, and the equipment to be maintained is scored for defects based on a defect metering model, so that the equipment problem can be quickly positioned. In the process of equipment maintenance, semantic analysis is carried out on historical defect information of equipment to be maintained, defect processing opinions are associated with the equipment to be maintained, after the equipment is maintained, a maintenance model is built based on the defect score and the defect processing opinions of the equipment to be maintained, and a maintenance plan of the equipment to be maintained is generated according to the maintenance model, so that defect detection can be automatically carried out on the positioned equipment, a corresponding solution scheme can be matched, and the equipment maintenance can be automatically and intelligently carried out.

Further, the step of using a defect metering model for the historical defect information of the equipment to be overhauled and scoring the defects of the equipment to be overhauled based on data analysis and data statistics comprises the following steps:

preprocessing the historical defect information of the equipment to be overhauled, and performing data analysis and statistics on the preprocessed data;

screening variables required by a defect metering model, establishing the defect metering model based on woe boxes, generating a defect scoring card according to a woe value, and scoring the overhaul equipment based on the defect scoring card.

From the above description, the score of the equipment to be overhauled is obtained through the defect metering model and the defect scoring card, and the states of the equipment and the defects thereof can be known.

Further, the scoring the equipment to be overhauled for defects based on data analysis and data statistics further comprises:

respectively scoring the equipment to be overhauled and the defects thereof, presetting a color value array, and calculating the color values of the equipment to be overhauled and the defects thereof:

defect color value = color value array defect scoring coefficient;

device color values = color value array device scoring coefficients;

and visually displaying the equipment to be overhauled and the defect information thereof based on the color values of the equipment to be overhauled and the defects thereof.

From the above description, the visual display of the defect score and the equipment score based on the color can more intuitively understand the equipment and the state of the defect thereof.

Further, performing semantic analysis on the historical defect information of the equipment to be overhauled, and associating corresponding defect processing opinions for the equipment to be overhauled includes:

performing text classification on the historical defect information of the equipment to be overhauled through semantic analysis, establishing a category tree according to category labels obtained through classification, and associating the historical information of the equipment to be overhauled based on the category tree;

and matching the historical information and the historical defect information of the equipment to be overhauled by using a convolutional neural network algorithm to obtain a corresponding defect processing opinion.

According to the description, the parallelization of the data and the theme model can be realized by adopting an intelligent semantic analysis algorithm, and the corresponding processing opinions can be obtained based on the defect information analysis.

Further, establishing a maintenance model based on the defect score and the defect treatment suggestion of the equipment to be maintained, generating a maintenance plan of the equipment to be maintained according to the maintenance model, and performing maintenance on the scanned equipment to be maintained by using the maintenance plan comprises:

analyzing according to the historical defect number, defect score and defect processing suggestion of the equipment to be overhauled, and establishing and storing an overhauling model;

and receiving the scanned equipment to be overhauled, carrying out overhaul accounting according to the overhaul model, periodically inquiring the overhaul plan of the equipment to be overhauled calculated by the overhaul model according to an accounting result, and actively reporting the overhaul requirement of the equipment to be overhauled.

According to the description, the maintenance plan which accords with the equipment defects can be found and reported automatically by analyzing the defect quantity, the defect score and the defect processing suggestion of the equipment to be maintained and constructing a model.

Referring to fig. 2, another embodiment of the present invention provides an intelligent overhaul terminal for devices based on big data and physical ID, including a memory, a processor, and a computer program stored in the memory and running on the processor, where the processor executes the computer program to implement the following steps:

According to the description, before the equipment maintenance is carried out, the equipment to be maintained obtained by scanning the equipment real object ID is obtained, the corresponding historical defect information is searched according to the information of the equipment to be maintained, and the equipment to be maintained is subjected to defect scoring based on the defect metering model, so that the equipment problem can be quickly positioned. In the process of equipment maintenance, semantic analysis is carried out on historical defect information of equipment to be maintained, defect processing opinions are associated with the equipment to be maintained, after the equipment is maintained, a maintenance model is built based on the defect score and the defect processing opinions of the equipment to be maintained, and a maintenance plan of the equipment to be maintained is generated according to the maintenance model, so that defect detection can be automatically carried out on the positioned equipment, a corresponding solution scheme can be matched, and the equipment maintenance can be automatically and intelligently carried out.

Further, the step of using a defect metering model for the historical defect information of the equipment to be overhauled and scoring the defect of the equipment to be overhauled based on data analysis and data statistics comprises the following steps:

Further, the step of scoring the defect of the equipment to be overhauled based on the data analysis and the data statistics further comprises the steps of:

defect color value = color value array defect score coefficient;

device color values = color value array device scoring coefficients;

The intelligent equipment maintenance method and terminal based on big data and physical ID are suitable for intelligent equipment maintenance, can quickly locate equipment problems and match corresponding solutions, and are described in the following through specific implementation modes:

example one

Referring to fig. 1 and 3, an intelligent overhaul method for equipment based on big data and physical ID includes the steps:

s1, acquiring equipment to be overhauled obtained by scanning equipment real object ID, and searching corresponding historical defect information according to the information of the equipment to be overhauled.

Specifically, after a maintainer holds a work ticket and enters the transformer substation, safety measures are started to be arranged. The method comprises the steps of scanning two-dimensional codes on each device of the current transformer substation by using an APP program, decoding the two-dimensional codes, simultaneously inquiring device and parameter information in a system, and carrying out association display on the device information and the transformer substation, so that the device to be overhauled obtained by scanning the APP is obtained.

And intelligently associating the previous defect condition of the equipment through the information of the equipment to be overhauled, and checking the information of the defect of the equipment.

And S2, using a defect metering model for the historical defect information of the equipment to be overhauled, and carrying out defect scoring on the equipment to be overhauled based on data analysis and data statistics.

Specifically, according to the historical condition of the defects, a defect metering model and a defect scoring card are adopted, each defect is scored through exploratory data analysis and descriptive statistics, and then the equipment is comprehensively scored, so that the equipment and the state of the defect thereof can be known more intuitively. The algorithm model has the advantages that the required data volume is small, the requirement on computer hardware is not high, the overall efficiency is improved by 93% compared with other grading schemes, and the requirement on hardware performance is reduced by 52%.

And S3, performing semantic analysis on the historical defect information of the equipment to be overhauled, and associating corresponding defect processing suggestions for the equipment to be overhauled.

Specifically, parallelization of data and a theme model is realized for equipment information through a parallel LDA (Latent Dirichlet Allocation) intelligent semantic analysis algorithm, and historical defect processing reports, equipment drawings, specifications, historical experience bases, auxiliary repair databases and other data and expert opinions are associated with the defect information through steps of data cleaning, word segmentation, part-of-speech tagging, named entity recognition, word vectors, syntactic semantic dependency analysis, similarity calculation, text classification tasks, text generation tasks and the like. Meanwhile, the expert system can provide processing opinions corresponding to the defects according to the characteristics of the defect information. And rapid query and view are supported at the APP terminal. Compared with a method for qualitatively determining defects by using multi-dimensional information, the parallel LDA algorithm mainly applied in the embodiment is remarkably improved by more than 20% in efficiency.

In some embodiments, after the equipment is overhauled, the overhaul test report is directly placed under a designated camera, the recorded information is automatically read through a self-defined Optical Character Recognition (OCR) technology, and an overhaul test report template is automatically filled in and stored in the PMS system, so that the equipment can be conveniently retrieved and read. Information formed at will after overhauling is carried out according to the equipment defects of the transformer substation, and after the information is identified through an OCR technology, standardized and digitized reports and contents can be conveniently and quickly formed, and the information can be conveniently called in the later period in the equipment overhauling process. The digital content is directly formed, the processes of report filling, copying and archiving are omitted, and quick inquiry and calling can be carried out in future use.

And S4, establishing a maintenance model based on the defect score and the defect processing suggestion of the equipment to be maintained, generating a maintenance plan of the equipment to be maintained according to the maintenance model, and using the maintenance plan to maintain the scanned equipment to be maintained.

Specifically, in the daily work after the equipment maintenance is completed, the system carries out model application on data modeling under limited information provided by a planning system, and establishes an equipment requirement model, wherein the model comprises three dimensions of defect quantity, defect severity and safety measures required by the defects. And automatically operating the data model through an Apriori association algorithm, finding a maintenance plan which accords with the defects of the equipment, and automatically reporting the maintenance plan. After scanning the real object ID of the equipment, the system reflects the maintenance requirement degree of the equipment according to the model accounting result: and when monitoring the plan meeting the working condition, the planning system actively reports the demand according to the demand model result. The method has the advantages that data walls among different systems are broken through, and the application is carried out under the condition that the respective safety of the existing systems is not influenced.

Therefore, the embodiment generates the corresponding detection scheme by scanning the device to be detected, greatly improves the matching efficiency and speed of defect discovery and relevant data breaking data barriers in the overhauling process, and greatly reduces the requirement on hardware; and the working efficiency is greatly improved, and the time cost is reduced. Taking a surveyor as an example, in the case of surveying a common line spacing example, a switch, a disconnecting link secondary drawing, an inspection manual, a CVT (capacitive voltage transformer) secondary wiring diagram, and the like are sufficiently prepared, and a man-hour of half a day to a day is required. After the maintenance method of the embodiment is adopted, the same line interval example inspection can be completed in only 1 to 2 hours, and the labor hour reduction effect is obvious.

In addition, the embodiment improves the plan coverage and avoids repeated power failure and repeated dispatching of personnel. Taking an interval example for inspection, arranging personnel to go to the correction and modification, needing to occupy one automobile, one driver and two maintainers, and working hours in one day. If the maintenance method of the embodiment is adopted, the condition can be avoided, and the maintenance method is combined with the maintenance work and only takes about 1 hour for modification.

Further, the embodiment can promote digital overhaul construction and reduce the use amount of paper materials. The drawings, specifications and test reports required on the site are usually counted by the shutter A4 paper, if hundreds of times of inspection are carried out in the power transformation center every year, tens of thousands of paper are consumed, and besides, the loss cost of the toner and the printer is increased. After the system is adopted, the working mode of digital overhaul can be realized, the cost is completely saved, 10000 sheets of paper and 4 ink powder boxes can be saved for A4 paper every year, and the number of the paper is reduced to 5000 yuan of RMB.

Example two

The difference between the present embodiment and the first embodiment is that a method for calculating a defect score is further defined, specifically:

screening variables required by a defect metering model, establishing the defect metering model based on woe boxes, generating a defect scoring card according to a woe value, and scoring the overhaul equipment based on the defect scoring card;

defect color value = color value array defect score coefficient;

device color value = color value array device scoring coefficient;

In the present embodiment, the defect risk measurement model includes a defect history rating, an overall equipment rating, and an overall line rating. The defect rating is composed of a series of rating models including a card a (device rating card), a card B (historical defect model), a card C (line model), and a card F (tracking model).

The typical defect scoring card model mainly comprises the following development processes:

1. data is acquired, including data for a history of device defects. The data includes dimensions of defects, including frequency of occurrence, level, impact, speed of defect removal, ease of defect removal, and the like.

2. The data preprocessing comprises the main work of data cleaning, missing value processing, abnormal value processing, data type conversion and the like, and the original data needs to be converted into the modeling data layer by layer.

3. EDA exploratory data analysis and descriptive statistics comprise the statistics of the size of the total data quantity, the defect level ratio, the data types, the variable missing rate, the variable frequency analysis histogram visualization, the box diagram visualization, the variable correlation visualization and the like.

4. And selecting variables, namely screening the variables which have the most obvious influence on the default state by using a statistical method and a machine learning method. There are many methods for variable selection, including iv, feature import, variance, etc. In addition, variables with too high a deletion rate, no business explanatory variables and no value variables should be deleted.

5. The main difficulties of model development and scoring card modeling are woe binning, fractional stretching and variable coefficient calculation. The woe sub-box is a difficult point in evaluating cards and needs abundant statistical knowledge and business experience. At present, the box separation algorithm is as many as 50, no unified standard exists, generally, automatic box separation of a machine is carried out, then box separation is carried out manually, finally, the final performance of a model is tested repeatedly, and the optimal box separation algorithm is selected preferentially.

6. And (3) verifying the model, verifying the distinguishing capability, the predicting capability, the stability, the sequencing capability and the like of the model, and forming a model evaluation report to draw a conclusion whether the model can be used. The model verification is not completed once, but is periodically verified after the model is modeled and before the model is online. Model development and maintenance is a cycle, not a single completion.

7. And the defect scoring card is generated according to the variable coefficient of the logistic regression and the woe value. The scoring card is convenient for service explanation, very stable and popular in the industry. The method is to convert the Logistic regression model probability score into a standard score form of 1-9 scores.

8. And establishing a scoring card model system, and establishing a computer automatic credit scoring system according to a credit scoring card method.

After the defects and the equipment are scored, the system visualizes the Data according to EDA (Exploratory Data Analysis) and descriptive statistics, marks the Data by using different colors, and can quickly find and check the corresponding equipment and the corresponding defects by using the equipment and the defect state colors in the list.

In the present embodiment, EDA and descriptive statistics include statistical population data size, defect level ratio, data type, variable missing rate, variable frequency analysis histogram visualization, box plot visualization, and variable correlation visualization, among others. The methods for analyzing exploratory data are commonly as follows: histogram, scatter plot, box plot, thermodynamic diagram, pairing plot.

Therefore, in the embodiment, according to the equipment of the substation and the defects thereof, deep statistics and analysis are performed, and classification is performed by combining the state and the severity of the defects. The system automatically scores each defect through EDA exploratory data analysis and descriptive statistics by adopting a defect metering model and a defect scoring card according to the historical condition of the defect, and then comprehensively scores the equipment so as to more intuitively know the equipment and the state of the defect. The algorithm model has the advantages that the required data volume is small, the requirement on computer hardware is not high, the overall efficiency is improved by 93% compared with other scoring schemes, and the requirement on hardware performance is reduced by 52%.

EXAMPLE III

The difference between the present embodiment and the first and second embodiments is that a method for generating a defect handling opinion is further defined, specifically:

In this embodiment, first, by analyzing the information of the device defect and its parameters, the big data is analyzed, and semantic combinations and frequency relationships are found and text classification is performed. Such as: the defect equipment is a 220kV imitative long III-way 249-unit combined electrical apparatus, and has the highest occurrence frequency of keywords of 220kV imitative long III-way 249, combined electrical apparatuses and the like. And storing the data through an intelligent semantic algorithm. Then, the information in the historical database, including drawings, specifications and the like, is also subjected to semantic splitting and storage.

Text classification is the most common text semantic analysis task, and is simple firstly, almost every text classification is done after being contacted with NLP, but is complex, and for a text classification task with hundreds of category labels, the accuracy rate of more than 90% is still a difficult matter. The text classification in this embodiment refers to a pan text classification, which includes a search classification, an advertisement classification, a page classification, a user classification, and the like.

Almost all machine learning methods can be used for text classification, but the best implementation of the following algorithm is adopted in the present embodiment, and the steps are as follows:

1. and establishing a classification system. On the one hand, the category system established by one person cannot cover all situations due to the limitation of the knowledge plane, and on the other hand, imbalance between categories can exist. Therefore, the relation between one class and another class is uncertain, which brings difficulties in use and optimization, so that the category system is set by manual identification and research and is processed based on the existing category system.

2. And establishing a category tree according to a certain hierarchical relationship according to the category label. And (4) classifying by using a hierarchical classifier, firstly training one classifier for the first-layer nodes, then training n classifiers for the second layer (n is the number of the nodes of the first layer), and so on. By using the hierarchical category tree, on one hand, a single model is simpler and more accurate, and on the other hand, the cross influence among category labels can be avoided.

Secondly, analyzing and discovering semantic combination and frequency relation through big data and splitting and storing the semantic combination and frequency relation through an intelligent semantic algorithm by information and main description of equipment defects. Then, the information in the historical database, including drawings, specifications and the like, is also subjected to semantic splitting and storage. The local optimal solution is deduced by performing semantic splitting and storage on the titles of a historical experience library and auxiliary repair data and performing parallelization LDA inference through an EM (Expectation-Maximization Algorithm) Algorithm. The association relation is stored through a program, and is displayed at the APP end through program calling, so that quick query and downloading viewing are supported.

1. The topic of each word is randomly initialized, and two frequency count matrixes are counted: a document topic count matrix N (t, d) for describing topic frequency distribution in each document; and a word topic count matrix N (w, t) for representing the frequency distribution of words under each topic.

2. And traversing the training corpus, resampling the theme corresponding to each word according to a probability formula, and updating the counts of N (t, d) and N (w, t).

3. The second step is repeated until the model converges.

Parallelization of the topic model, not only the topic model, can be mainly explained from two perspectives: data parallel and model parallel.

The data is parallel to the model and can be described as a chessboard, the rows of the chessboard are divided according to the data, and the columns of the chessboard are divided according to the model. The parallelization of LDA is to divide the originally huge matrix which cannot be stored in a single machine into different machines through the division, so that each machine can store the parameters in the memory. Then, each word vector is relatively independently calculated, and model data are synchronized at intervals in the calculation process according to certain strategies.

In the text analysis model, vectors are used to describe a word "word vector" for a word. The application of word vectors makes the model simpler and truly predicted using context. The basic principle is as follows: the application word vector of the word vector can mine the relation between words, apply the word vector as a characteristic to other machine learning tasks, can be applied to machine translation, can utilize the characteristic, draw advantages such as hierarchical relation between words.

And adding the word vectors corresponding to all the words to form a phrase vector and a sentence vector. Analysis is then performed on the phrase vectors and sentence vectors. And adding an ID in the training process, namely each sentence in the training corpus has a unique ID. And in the prediction stage, a paragraph id is newly distributed to the sentence to be predicted, the parameters obtained in the training stage are kept unchanged by the word vector and the parameters of the output layer, and the sentence to be predicted is trained by reusing gradient descent. And after convergence, obtaining the vector of the sentence to be predicted.

And then, after semantic splitting is carried out according to the equipment type information in the defect library, the equipment type information is positioned under the equipment type corresponding to the expert library. And performing semantic splitting on the information in the defect content in the defect library to extract two columns of information with the components and the problem description in the expert library. And applying a convolutional neural network algorithm based on CNN (convolutional neural network) to the text, and further achieving all expert processing opinions matching the equipment and the defects so as to quickly check related opinions for repairing.

Based on CNN (Convolutional Neural Networks), the method can be used for text classification, emotion analysis, ontology classification and the like. Tasks such as traditional text classification are generally based on word senses or on feature extraction of words themselves, and such methods generally require domain knowledge and artificial features. The method is similar by using the CNN, but the CNN model is generally based on the original text, and the input of the CNN model can be a word set, a word vector or a simple character. Compared to conventional approaches, CNN does not require excessive human characterization.

And taking the word set as input, and classifying the text by using CNN. The CNN is divided into four layers, the first layer is a word vector layer, each word in the file is mapped to a word vector space, and n words are mapped after being assumed to be k-dimensional, namely an n-x-k-dimensional image is generated; the second layer is a convolution layer, a plurality of filters act on the word vector layer, and different filters generate different characteristic graphs; the third layer is a convergence layer, and the maximum value of each feature graph is taken, so that the operation can process variable-length documents, and the output of the third layer only depends on the number of filters; the fourth layer is a fully-connected highest soft layer, and the output is the probability for each category. In addition, the input layer can have two channels, wherein one channel adopts word vectors which are trained by using a parallel LDA algorithm in advance, and the word vectors of the other channel can be adjusted in the training process through a reverse algorithm. The result of this is: of the 7 triage tasks currently in common use, 4 achieved the desired results, and the other 3 performed near best.

From the above description, in the process of acquiring and associating relevant data of equipment and defects thereof, the system adopts technologies of intelligent semantic algorithms such as big data analysis, text classification method based on text semantic analysis, parallelized LDA algorithm, CNN algorithm based on convolutional network, and the like, performs semantic statistics, splitting, matching and storage, and performs storage, calling and display by using methods of a database and the system. Thereby, a logical association between different data repository information is achieved.

Example four

The difference between this embodiment and the first to third embodiments is that a method for generating a maintenance plan is further defined, specifically:

In this embodiment, the following are specifically mentioned:

1. the system establishes an equipment requirement association model by adopting an association rule Apriori algorithm according to three dimensions of the number of the defects of the equipment, the severity of the defects and safety measures required by the defects, and applies data modeling.

2. And after scanning the equipment object ID, the system performs result accounting according to the model. And when the planning system monitors a plan meeting the working conditions, actively reporting the maintenance requirement degree of the equipment according to the requirement model accounting result.

And establishing an equipment requirement association model by adopting an association rule Apriori algorithm according to three dimensions of the number of the defects, the severity of the defects and safety measures required by the defects of the equipment through data analysis. The system automatically invokes a matching application on the data modeling. And after scanning the equipment real object ID, the system automatically matches the model accounting result according to the inquired equipment information. If a plan meeting working conditions is monitored in the planning system, the maintenance requirement degree of the equipment is reflected and actively matched according to the requirement model result, and the plan requirement matched with the model is automatically displayed and reported when defects are checked.

According to the description, the defect information (the main defect quantity, the defect severity and the required safety measure for the defects) of the equipment history of the transformer substation is analyzed, and a set of unique equipment requirement model is formed by adopting an association rule Apriori algorithm. Through the model, the system can automatically match and automatically associate the maintenance plan of the current defects of the equipment according to the equipment information, so that active reporting is performed.

Where association rules are unsupervised machine learning methods used for knowledge discovery, rather than prediction. The learner of the association rules need not label the training data in advance because unsupervised learning does not train this step. Meanwhile, the model has the defect that the model evaluation of the association rule learner is difficult, and whether the result is reasonable can be generally observed through business experience.

The Apriori principle is: a certain set of items is frequent, so all its subsets are also frequent. Apriori's principle can help reduce the computational load more often it is its inverse negative proposition, i.e. if a set of terms is infrequent, then all of its supersets are also infrequent, called the inverse monotonicity of the set of terms, closure down.

The method comprises the following specific steps:

1. and generating a candidate 1-item set C1, calculating the support degree, and generating a frequent 1-item set L1 according to the minimum support degree.

The item set is called an item set, the number of elements is the length of the item set, and the item set with the length of k is called a k-item set.

Item set support is used to describe the importance of X, for item set X, count is the number of transactions containing X in transaction set D, and the support of item set X is the probability of occurrence of item set X:

2. and generating a candidate 2-item set C2, calculating the support degree, and generating a frequent 2-item set L2 according to the minimum support degree.

3. And generating a candidate 3-item set C3, calculating the support degree, and generating a frequent 3-item set L3 according to the minimum support degree.

4. And generating an association rule. The simplest method is to list all non-empty proper subsets of each frequent item set, take any two of the frequent item sets as LHS (Left Hand Side) and RHS (Right Hand Side), form association rules, calculate the confidence of each association rule, and delete weak rules.

The system automatically triggers the model after scanning the real object ID of the equipment, periodically inquires the plan which is calculated by the model and is monitored by the associated planning system to meet the working condition according to the result storage condition of model accounting, and actively reports the maintenance requirement degree of the equipment according to the requirement model accounting result.

From the above description, in the daily operation of the equipment, the purpose of automatically and regularly carrying out the planned detection of the equipment is realized by analyzing and modeling the number of the defects, the severity of the defects and the safety measures required by the defects. If the maintenance plan which accords with the equipment defects is found through model operation, the maintenance plan can be automatically reported, so that the labor input for observation and analysis in the daily operation of the equipment is greatly saved.

EXAMPLE five

Referring to fig. 2, an intelligent overhaul terminal 1 for a device based on big data and physical ID includes a memory 2, a processor 3, and a computer program stored in the memory 2 and capable of running on the processor 3, where the processor 3 implements each step of an intelligent overhaul method for a device based on big data and physical ID in any one of the first to fourth embodiments when executing the computer program.

In summary, according to the intelligent overhaul method and the terminal for the equipment based on the big data and the physical ID provided by the invention, before the equipment overhaul is performed, the system can automatically decode and query the obtained equipment information and the defect information of the equipment by scanning the physical ID code of the equipment. And (3) scoring each defect by adopting a defect metering model and a defect scoring card through EDA exploratory data analysis and descriptive statistics so as to more intuitively know the equipment and the state of the defect. In the equipment maintenance process, the system parallels an LDA intelligent semantic analysis algorithm to realize parallelization of data and a theme model, and automatically inquires information such as drawings, specifications, historical experience bases, auxiliary repair databases and expert opinions of associated equipment and defects thereof through steps of data cleaning, word segmentation, part of speech tagging, named entity recognition, word vectors, syntax semantic dependency analysis, similarity calculation, text classification tasks and text generation tasks, so that maintenance can be quickly carried out by referring. After the equipment maintenance is completed, the system can quickly generate and store a maintenance test report through an OCR technology. In the daily work after the equipment maintenance is finished, the system automatically runs the data model through the running data model, automatically runs the data model through an Apriori association algorithm under the limited information provided by the planning system, finds the maintenance plan which accords with the equipment defects, and automatically reports the maintenance plan. Through the end operation to among the overhaul of the equipments in-process APP, the efficiency of the overhaul of the equipments work that can be very big also can improve the plan coverage simultaneously, avoids repeated power failure and the repeated dispatch of personnel to impel the digital maintenance construction of overhaul of the equipments process and daily management.

The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent changes made by using the contents of the present specification and the drawings, or applied directly or indirectly to the related technical fields, are included in the scope of the present invention.

Claims

1. An intelligent equipment maintenance method based on big data and physical ID is characterized by comprising the following steps:

2. The intelligent overhaul method for the equipment based on the big data and the physical ID as claimed in claim 1, wherein the using a defect metering model for historical defect information of the equipment to be overhauled, and the defect scoring for the equipment to be overhauled based on data analysis and data statistics comprises:

3. The intelligent overhaul method for the equipment based on the big data and the physical ID as claimed in claim 1, wherein the defect scoring for the equipment to be overhauled based on the data analysis and the data statistics further comprises:

defect color value = color value array defect score coefficient;

device color values = color value array device scoring coefficients;

4. The intelligent overhaul method of the large data and physical ID based equipment according to claim 1, wherein performing semantic analysis on historical defect information of the equipment to be overhauled and associating corresponding defect handling ideas for the equipment to be overhauled comprises:

5. The intelligent overhaul method for the equipment based on the big data and the physical ID as claimed in claim 1, wherein a overhaul model is built based on the defect score and the defect treatment suggestion of the equipment to be overhauled, an overhaul plan for the equipment to be overhauled is generated according to the overhaul model, and the overhaul of the scanned equipment to be overhauled by using the overhaul plan comprises:

6. An intelligent overhaul terminal of equipment based on big data and entity ID, comprising a memory, a processor and a computer program stored on the memory and capable of running on the processor, characterized in that the processor implements the following steps when executing the computer program:

7. The intelligent overhaul terminal based on big data and physical ID of claim 6, wherein the using a defect metering model for historical defect information of the equipment to be overhauled and the scoring of the defect of the equipment to be overhauled based on data analysis and data statistics comprises:

8. The intelligent overhaul terminal based on big data and physical ID of claim 6, wherein the fault scoring of the equipment to be overhauled based on data analysis and data statistics further comprises:

defect color value = color value array defect scoring coefficient;

device color values = color value array device scoring coefficients;

9. The intelligent overhaul terminal based on big data and physical ID of claim 6, wherein performing semantic analysis on historical defect information of the equipment to be overhauled, and associating corresponding defect handling ideas for the equipment to be overhauled comprises:

10. The intelligent overhaul terminal based on big data and physical ID of claim 6, wherein a overhaul model is built based on the defect score and the defect treatment suggestion of the equipment to be overhauled, an overhaul plan of the equipment to be overhauled is generated according to the overhaul model, and the overhaul of the scanned equipment to be overhauled by using the overhaul plan comprises: