CN110909895A

CN110909895A - Early warning method and system based on special equipment historical periodic inspection report

Info

Publication number: CN110909895A
Application number: CN201911134951.7A
Authority: CN
Inventors: 裴明丽; 周源; 严伟; 王启龙
Original assignee: Section Big Country Wound Software Inc Co
Current assignee: Section Big Country Wound Software Inc Co
Priority date: 2019-11-19
Filing date: 2019-11-19
Publication date: 2020-03-24

Abstract

The invention discloses an early warning method and system based on a special equipment historical periodic inspection report, and relates to the technical field of data mining and application. The invention comprises the following steps: step S1: acquiring original record data in the historical inspection report of the special equipment through an inspection template; step S2: preprocessing the obtained original recorded data, constructing a training data set, a verification data set and a test data set broad table, and obtaining analysis data; step S3: establishing a fault early warning model of a periodic inspection report of the special equipment by adopting a supervised decision tree classification algorithm; step S4: and measuring the accuracy of the prediction result of the early warning model by using various evaluation indexes. According to the invention, basic information data of special equipment and inspection data in historical periodic inspection reports are mined through a data mining and analyzing technology, and a fault early warning model is established by utilizing a decision tree classification algorithm, so that the working efficiency of inspection departments is improved, and the personal and property safety of social personnel is ensured.

Description

Early warning method and system based on special equipment historical periodic inspection report

Technical Field

The invention belongs to the technical field of data mining and application, and particularly relates to an early warning method and system based on a special equipment historical periodic inspection report.

Background

With the rapid development of economy, the physical living standard of people is continuously improved, in order to meet the living needs of people, the number of special equipment in buildings such as houses, companies, markets and the like is continuously increased, but the equipment is frequently operated and is not maintained in place along with the lapse of time, so that the equipment is gradually aged, the performance is continuously reduced, accidents can happen at any time, personnel injury and property loss are caused, and the social attention is widely paid.

Therefore, the periodic inspection of special equipment is a crucial detection link, and it is the responsibility of the inspector to accurately and completely record the periodic inspection report and report the inspection result in time. When the inspection is carried out regularly, different inspection results are generated due to different factors such as different batches of products and different models of products, a large amount of manpower and material resources are consumed, a large amount of data information is accumulated along with the lapse of time, if the information data is not utilized, the information data becomes waste data and dead data, and more deep and meaningful data cannot be generated. The basic information data of the special equipment and the inspection data in the historical periodic inspection report are mined through a data mining and analyzing technology, whether the next inspection result can pass or not can be predicted, so that measures can be taken in advance to prevent accidents from happening, social stability is maintained, threats brought by the accidents to the health and property safety of people are reduced, meanwhile, the detection efficiency is improved, and a better special equipment detection safety system is formed.

Disclosure of Invention

The invention aims to provide an early warning method and system based on a historical periodic inspection report of special equipment.

In order to solve the technical problems, the invention is realized by the following technical scheme:

the invention relates to an early warning method based on a special equipment historical periodic inspection report, which comprises the following steps:

step S1: acquiring original record data in a special equipment past inspection report through an inspection template, acquiring historical inspection record data of the equipment, reading an equipment parameter table and acquiring basic information data of the equipment;

step S2: preprocessing the obtained original recorded data, constructing a training data set, a verification data set and a test data set broad table, and obtaining analysis data;

step S3: establishing a fault early warning model of a periodic inspection report of the special equipment by adopting a supervised decision tree classification algorithm;

step S4: measuring the accuracy of the prediction result of the early warning model by using various evaluation indexes;

step S5: the model is deployed and applied to actual production and living environments;

step S6: and feeding back the early warning fault condition, and informing a supervision and maintenance unit to carry out detection and maintenance.

Preferably, in step S1, the original record data collected by the inspection template is the historical periodic inspection report data of the special equipment, and when the same equipment is inspected, at least the inspection data in the past several years needs to be collected.

Preferably, in step S2, the data preprocessing operation specifically includes a data cleaning unit, a feature screening unit, and a feature constructing unit; the data preprocessing is used for searching effective fields to construct a training set, a verification set and a test set width table.

Preferably, the wide table is established by adopting a sliding window method for the training set and the verification set wide table, namely, the device codes of three adjacent months of M-1 and M, M +1 are used as a label set, and then the label set is associated with device basic data information and device inspection record data to obtain the training set, the verification set and the test set wide table.

Preferably, in the step S4, the multiple evaluation index metrics are optimized for the fault-warning model by using precision, recall and F1-Score.

The invention relates to an early warning system based on historical periodic inspection reports of special equipment, which comprises a data acquisition and preprocessing module, a mining and early warning module, a model deployment and application module and a feedback and supervision module, wherein the data acquisition and preprocessing module is used for acquiring and preprocessing data;

the data acquisition and preprocessing module, the mining and early warning module, the model deployment and application module and the feedback and supervision module are sequentially connected;

the data acquisition and preprocessing module is used for reading a detection report through a detection template, obtaining fixed information parameter data and detection related data, and cleaning, screening and constructing the data to further obtain usable and analyzable data;

the mining and early warning module is used for constructing a training data set, a verification data set and a test data set, training the training data set by adopting a supervised decision tree classification algorithm-Lightgbm algorithm, verifying the training data set by using the verification data set, evaluating the model by adopting precision ratio, recall ratio and F1-Score index, and finally predicting faults in the test data set;

the feedback and supervision module: the system is used for feeding back early warning fault conditions, informing a supervisor and carrying out maintenance in advance;

the model deployment and application module is used for deploying models and applying the models to actual production life, and brings more convenience and value to life.

The invention has the following beneficial effects:

(1) according to the invention, the original record data in the past inspection report of the special equipment is acquired through the inspection template, so that the potential useful information of the equipment can be mined and analyzed, the working efficiency of an inspection department is improved, and the personal and property safety of social personnel is ensured.

(2) According to the invention, the data in the inspection report is converted into structured, simple and understandable data by a data preprocessing method, and indexes which have value significance and can be analyzed are screened and constructed by utilizing a characteristic screening and characteristic constructing method, so that later-stage modeling is facilitated, the detection efficiency is improved, and a large amount of manpower and material resources are avoided.

(3) The method adopts a sliding window method to construct a training data set and a verification data set broad table, utilizes a common supervised decision tree classification algorithm in data mining and analysis to carry out modeling, adopts various measurement methods to evaluate the modeling effect, and makes the model more suitable for production and actual life.

Of course, it is not necessary for any product in which the invention is practiced to achieve all of the above-described advantages at the same time.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a diagram of steps of an early warning method based on historical periodic inspection reports of special equipment according to the present invention;

FIG. 2 is a schematic structural diagram of an early warning system based on historical periodic inspection reports of special equipment according to the present invention;

FIG. 3 is a feature set, tagset data source construction diagram of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

Referring to fig. 1, the present invention is a method for warning based on historical periodic inspection reports of special equipment, including the following steps:

step S3: establishing a fault early warning model of a periodic inspection report of the special equipment by adopting a supervised decision tree classification algorithm Lightgbm;

In step S1, since the device basic information data refers to the self-contained attribute of the device, the value thereof is unique, the device inspection record data is dynamically changed, and each inspection may be different, so when the original record data collected by the inspection template is used as the periodic inspection report data of the special device, the inspection data in the last several years needs to be collected at least when the same device is inspected, which is convenient for data preprocessing and easy to achieve modeling effect.

The fields specifically included in the device basic information data and the device history check record data are shown in tables 1 and 2 below:

name of field	Field coding
		Device code	SBDM
Field value	ZDZ
		Class of devices	SBLB_MC
Variety of equipment	SBPZ_MC
		Date of manufacture	ZZRQ
Name of manufacturing Unit	ZZDW_MC
		Name of unit of use	SYDW_MC
Site of use of the apparatus	SBSZDD
		Date of transformation	GZRQ
Name of reforming unit	GZDW_MC
		Name of maintenance unit	WBDWMC
Rated load capacity	P30002004
		Rated speed	P30002003
Control mode	P30001001

Table 1 is a data table of basic information of the device

Name of field	Field coding
		Device code	SBDM
Report numbering	BGBH
		Test conclusion	JYJL
Number of qualified items	HG
		Number of unqualified items	BHG
Nothing here	WCX

Table 2 shows a data table of the history of the equipment

In step S2, the data preprocessing operation specifically includes a data cleaning unit, a feature screening unit, and a feature constructing unit; the data preprocessing is used for searching effective fields to construct a training set, a verification set and a test set width table, so that the data can be accurate and standard;

the data cleansing unit generally includes checking data consistency, processing outliers and duplicate values.

i. For data consistency: checking data consistency conditions related to date fields, such as whether field forms and types such as inspection date, manufacturing date, modification date and the like are consistent or not, and converting inconsistent fields into a uniform format; indexes such as rated load capacity, rated speed and the like are processed, and unit inconsistency is converted into consistency;

ii, for abnormal values, such as descriptive results of control mode field values with # characters and undetected characters, filling the abnormal values into 0 according to business understanding and field characteristics, or deleting the abnormal values, or filling the abnormal values according to methods of average number, mode, median and the like;

and iii, aiming at the repeated values, if a plurality of equipment codes correspond to a plurality of using unit names, sorting in a descending order according to the inspection date, and taking a latest record.

The feature filtering unit applies relevance analysis to remove field features without relevance features, for example, the field value is null, or the field value has only one value, or each field value is different. The method is used for deleting the records with null field values in the inspection records, such as deleting the names of maintenance units, main safety management personnel, inspectors and fields of grounding connection, electrical insulation and the like in inspection items;

the characteristic construction unit is generally used for constructing an equipment basic information data table and an equipment historical check record data table;

① for device basic information data sheet:

i. firstly, counting the distribution condition of values of a control mode field, then cleaning the values to be common values, and then carrying out one _ hot operation on the values;

performing one _ hot operation on the rated load capacity, the rated speed, the equipment variety, the equipment type and the equipment use place; converting the format of the manufacturing date and the transformation date, constructing a manufacturing year field and a transformation year field of the equipment according to the year value, and performing one _ hot operation on the manufacturing year field and the transformation year field;

iii, counting and ranking the using units, the transforming units and the manufacturing units, wherein the original unit names of the 150 units before ranking are used, the unit names of the 150 units after ranking are replaced by others, and then performing one _ hot operation;

② check records data sheet for device history:

i. the test items in each device comprise a plurality of different small test items and small test conclusions, the qualified items, the unqualified items and the quantity of the unqualified items in the test items are firstly counted, and then six statistics are carried out: maximum, minimum, mean, standard deviation, count, and median;

count the number of historical checks per device code.

Referring to fig. 3, the training set and verification set width table is created by using a sliding window method, that is, device codes of three adjacent months, namely M-1 and M, M +1, are used as a label set, and then are associated with device basic data information and device inspection record data to obtain a training set, a verification set and a test set width table.

For example, when we need to predict whether the label of 5 months in 2019 is qualified, data of two months of inspection in 2019 of the equipment inspection date should be extracted as a training set and a verification set respectively, and the inspection conclusion is taken as the label, and the specific process is as follows:

training set width table: taking the equipment code and label with the inspection date of 3 months as a label set, and then associating a relevant table through an equipment code field, namely associating an equipment basic information data table, an equipment history inspection record data table and an index field expanded in the steps as a training set;

verification set width table: taking the equipment code and label with the check date of 4 months as a label set, and then associating as a verification set in the same way as the step of the training set width;

test set width table: taking the equipment code with the check date of 5 months, and then associating as described in the step of training set width table to be used as a test set;

as shown in table 3 below:

name of field	Field coding
		Device code	SBDM
Test conclusion	JYJL

Table 3 is a tag set table

In step S3, the algorithm modeling is to train the training set broad table by using the supervised decision tree classification algorithm Lightgbm to obtain a fault early warning model, and then verify the obtained fault early warning model by using the verification set broad table, where the verification set confusion matrix is shown in table 4 below:

TABLE 4 validation set confusion matrix

In step S4, the multiple evaluation index metrics are optimized for the fault early warning model by using precision ratio, recall ratio, and F1-Score, so as to obtain an optimal model and ensure accuracy of a prediction result; outputting a corresponding result according to the fault early warning model:

the Recall ratio Recall obtained by calculation is as follows:

the Precision calculated is:

the calculated F1-Score was:

the calculation results show that the prediction result of the special equipment inspection conclusion is acceptable, and the results also show that the precision ratio and the recall ratio F1-Score of the model can be accurately judged by pertinently adjusting the parameters and the construction word fields of the early warning model, so that the model is higher in accuracy.

Referring to fig. 2, the present invention is an early warning system based on historical periodic inspection reports of special equipment, which includes a data collecting and preprocessing module, a mining and early warning module, a model deploying and applying module, and a feedback and monitoring module;

the acquisition and preprocessing data module is used for reading the inspection report through the inspection template, obtaining fixed information parameter data and inspection related data, and cleaning, screening and constructing the data to further obtain usable and analyzable data;

a feedback and supervision module: the system is used for feeding back early warning fault conditions, informing a supervisor and carrying out maintenance in advance;

and the model deployment and application module is used for deploying models and applying the models to actual production life, and brings more convenience and value to life.

A large number of data experiments prove that: the method comprises the steps of acquiring basic information data of equipment and historical inspection record information data of the equipment, converting the data in an inspection report into structured and easily-understood data by using a data preprocessing method so as to perform feature screening and feature construction, constructing a training set, a testing set and a verification set data width table by using a sliding window method according to constructed fields, the basic information data of the equipment and the historical inspection record data, modeling by using a supervised decision tree classification algorithm Lightgbm, and measuring and evaluating a model by using a plurality of indexes so as to obtain higher accuracy on an inspection result.

It should be noted that, in the above system embodiment, each included unit is only divided according to functional logic, but is not limited to the above division as long as the corresponding function can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.

In addition, it is understood by those skilled in the art that all or part of the steps in the method for implementing the embodiments described above may be implemented by a program instructing associated hardware, and the corresponding program may be stored in a computer-readable storage medium.

The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An early warning method based on historical periodic inspection reports of special equipment is characterized by comprising the following steps:

2. The early warning method based on historical periodic inspection reports of special equipment as claimed in claim 1, wherein in step S1, the original recorded data collected by the inspection template is historical periodic inspection report data of the special equipment, and when the same equipment is inspected, at least inspection data in past years needs to be collected.

3. The early warning method based on the historical periodic inspection report of the special equipment as claimed in claim 1, wherein in the step S2, the data preprocessing operation specifically comprises a data cleaning unit, a feature screening unit and a feature constructing unit; the data preprocessing is used for searching effective fields to construct a training set, a verification set and a test set width table.

4. The early warning method based on the historical periodic inspection reports of the special equipment as claimed in claim 3, wherein the wide table of the training set and the wide table of the verification set are established by a sliding window method, namely, the equipment codes of three adjacent months of M-1 and M, M +1 are used as label sets, and then are associated with equipment basic data information and equipment inspection record data to obtain the wide table of the training set, the wide table of the verification set and the wide table of the test set.

5. The warning method based on the historical periodic inspection report of the special equipment as claimed in claim 1, wherein in step S3, the trained fault warning model adopts a supervised decision tree classification algorithm Lightgbm, and the obtained fault warning model is verified by using a verification collection broad table.

6. The warning method based on historical periodic inspection reports of special equipment as claimed in claim 1, wherein in step S4, the multiple evaluation index metrics are optimized for the fault warning model by using precision ratio, recall ratio and F1-Score.

7. An early warning system based on historical periodic inspection reports of special equipment comprises a data acquisition and preprocessing module, a mining and early warning module, a model deployment and application module and a feedback and supervision module; the method is characterized in that:

the model deployment and application module is used for deploying models and applying the models to actual production and life.