CN106054857B - Maintenance decision tree/word vector-based fault remote diagnosis platform - Google Patents

Maintenance decision tree/word vector-based fault remote diagnosis platform Download PDF

Info

Publication number
CN106054857B
CN106054857B CN201610364696.5A CN201610364696A CN106054857B CN 106054857 B CN106054857 B CN 106054857B CN 201610364696 A CN201610364696 A CN 201610364696A CN 106054857 B CN106054857 B CN 106054857B
Authority
CN
China
Prior art keywords
cell
fault
word
unit
words
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201610364696.5A
Other languages
Chinese (zh)
Other versions
CN106054857A (en
Inventor
田雨农
张祥
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Dalian Roiland Technology Co Ltd
Original Assignee
Dalian Roiland Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Dalian Roiland Technology Co Ltd filed Critical Dalian Roiland Technology Co Ltd
Priority to CN201610364696.5A priority Critical patent/CN106054857B/en
Publication of CN106054857A publication Critical patent/CN106054857A/en
Application granted granted Critical
Publication of CN106054857B publication Critical patent/CN106054857B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00—Testing or monitoring of control systems or parts thereof
    • G05B23/02—Electric testing or monitoring
    • G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
    • G05B23/0208—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterized by the configuration of the monitoring system

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention relates to the technical field of fault remote diagnosis, in particular to a fault remote diagnosis platform based on a maintenance decision tree/word vector. The invention relates to a fault remote diagnosis platform, which comprises: the fault information receiving unit is used for receiving the fault description statement T; the fault information analysis unit is used for cutting words in the cell word bank to search the cell words and arranging the cell words; and the fault information solving unit is used for providing a maintenance solution through decision classification of the decision tree model. The invention has the beneficial effects that: the fault of the equipment can be remotely and quickly identified, and an effective maintenance solution can be accurately provided. The detection standards are unified, and errors caused by manual judgment are avoided. And a large amount of manpower and material resources are saved.

Description

Maintenance decision tree/word vector-based fault remote diagnosis platform
Technical Field
The invention relates to the technical field of fault remote diagnosis, in particular to a fault remote diagnosis platform based on a maintenance decision tree/word vector.
Background
The traditional fault diagnosis in the equipment maintenance industry uses an experience mode of 'watching, hearing, asking and cutting'.
The general diagnostic procedure is:
1. the trouble lamp of the equipment is lighted or the user feels that the equipment has problems, such as abnormal sound, peculiar smell, shaking and the like, namely 'finding a problem and entering a store'.
2. The fault code and the fault description are detected by a professional fault diagnosis instrument at a maintenance point, namely, the detection problem is detected.
3. The service technician performs on-site diagnostics and presents a solution, i.e., a "solution decision," by looking at the trouble code and trouble description, in conjunction with years of service experience.
4. The spare part person makes an order to give the specific spare part number, the name of the spare part and the price of the spare part, namely 'the spare part is made an order'.
5. The service advisor orders the number of work items, name of work items, hours, and time charges, i.e., "work item order.
In the prior art, a user needs to find a problem by means of experience when diagnosing equipment faults. After confirming that there is indeed a problem that needs to be solved by maintenance, the user needs to transport the equipment to a maintenance site to detect the fault, which is time consuming and labor intensive. During fault detection, a maintenance technician obtains a maintenance scheme by combining own experience, and gives a spare part number, a spare part name and a spare part price, and the process is highly subjective and lacks of unified standards. Then the inspected person and the service consultant give out different documents respectively, which is time-consuming and labor-consuming.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a fault diagnosis platform, which is used for constructing a maintenance cell word bank from fault description to word vector decomposition; by means of word segmentation retrieval of fault description in a cell word bank and according to a syntactic rule established by cell word weight, a maintenance decision tree from a decision process to a decision suggestion to a maintenance solution is realized, finally, remote fault diagnosis is realized, and labor cost and spare part cost are evaluated.
The technical problem to be solved by the invention is realized by adopting the following technical scheme: a fault remote diagnosis platform comprising: the fault information receiving unit is used for receiving the fault description statement T; the fault information analysis unit is used for cutting words in the cell word bank to search the cell words and arranging the cell words; and the fault information solving unit is used for providing a maintenance solution through decision classification of the decision tree model.
Preferably, the failure information analysis unit comprises a failure description statement decomposition unit, a retrieval unit, a calculation unit and an arrangement unit; the sentence decomposition unit is used for cutting the words of the fault description sentence T in the cell word stock; the retrieval unit is used for retrieving in the cell word bank; the calculating unit is used for calculating the weight of each cell word; the arrangement unit is used for arranging the cell words according to the syntactic rule.
Preferably, the search unit performs word segmentation search.
Preferably, the method for calculating the weight of each cell word by the calculating unit is as follows: s1, calculating chi-square statistic of each cell word; s2, the ith cell word with the highest chi-square statistic score is taken, and the frequency of occurrence of the ith cell word in the jth fault description is calculated.
Preferably, in the step S1, the chi-square statistic score is calculated by:
weight＝round(10×(1+lg(tfij))/(1+lg(lj)))；
wherein tf isijThe number of times the ith cellular word appeared in the jth failure description, ljThe length of the description for the jth fault.
Preferably, the fault information solution unit provides a maintenance solution in combination with the component measurement values.
The invention has the advantages and positive effects that: the fault of the equipment can be remotely and quickly identified, and an effective maintenance solution can be accurately provided. The detection standards are unified, and errors caused by manual judgment are avoided. And a large amount of manpower and material resources are saved.
Drawings
FIG. 1 is a diagram of the construction of syntactic rules according to the present invention;
FIG. 2 is a decision tree analysis classification model 1;
FIG. 3 is a decision tree analysis classification model 2.
Detailed Description
The present invention will be described in more detail below with reference to the accompanying drawings by way of specific embodiments. The following examples are illustrative only, not limiting, and are not intended to limit the scope of the invention.
The invention achieves the purpose of converting the problem of natural language understanding into the problem of machine self-learning by digitalizing the natural language.
(1) Constructing a base corpus
The method of creating the base corpus is to give a string whose natural language probability is P (w1, w2, w3, …, wn), w1 to wn being the respective cell words of this sentence in turn.
The model is based on the assumption that the occurrence of the nth word is only related to the first n-1 words and not to any other words, and that the probability of a complete sentence is the product of the probabilities of occurrence of the respective participles.
For a fault description T, its probability calculation formula is:
P(T)＝P(w1,w2,w3,…,wn)
＝P(w1)×P(w2|w1)×P(w3|w1,w2)×…×P(wn|w1,w2,…,wn-1)
≈P(w1)×P(w2|w1)×P(w3|w2)…P(wn|wn-1)
(2) and (3) decomposing the word vector: the word vector is a sentence that is segmented into a number of cell words. Each fault description T may be represented as an n-dimensional word vector, T ═ w1,w2,...,wn) Wherein w is1To wnFor each cell word contained in each word vector.
Taking the automobile fault as an example:
the fault is described as:
the fault description statement is expressed by a word vector, and the fault description statement can be expressed by a word vector (w)1,w2,w3,…,wn). Wherein, w1Being a camshaft, w2Is a position, w3Is a cylinder, w4Is a positive timing, w5Is excessive. Selecting each cell word as a characteristic item, wherein wiThe ith feature item is represented.
Calculating chi-square statistic of each cell word, and taking the ith cell word with the highest score in the training set as a representative word to find out a word vector (w)1,w2,w3,…,wn) The least relevant cell word.
wiThe formula for calculating the weight of (a) is:
weight＝round((10×(1+lg(tfij))/(1+lg(lj))),3)；
wherein tf isijFor the number of times the ith word vector appears in the jth fault description, ljThe length of the description for the jth fault. The Round function is rounded to the designated decimal place, which is the third place after the decimal point in the above equation.
The feature expression of the fault descriptor vector after the weight is calculated is as follows:
vector participle Weight of
Cam shaft 0.286
Position of 0.191
Cylinder 0.048
Timing of 0.048
Transition of 0.048
Example 1
(1) Search arrangement cell word:
as shown in fig. 1, when the user inputs the fault description T, the fault description T is first subjected to word vector decomposition to segment the complete sentence into a combination of a plurality of cell words. Then, word segmentation search is carried out in the cell word bank aiming at each cell word. If so, calculating the weight of the cell word; if not, storing the cell words which are not searched in the newly added industry cell word bank, searching the rest cell words, and calculating the probability.
And calculating weights according to the cell words searched in the cell word bank, and arranging the cell words from large to small according to the weights.
(2) And (3) analysis results:
taking the automobile fault as an example:
the fault is described as:
and (3) searching each cell word in the cell word bank, calculating the weight, and arranging the cell words from large to small according to the weight:
using discriminant analysis:
the word vector sequence of the fault description T is T ═ (w1, w2, w3,.., wn), and the sequence of each cellular word in the fault description T after the weight is calculated is S ═ (S1, S2, S3.., sn). The calculation formula of the word vector classification matching maximum probability is as follows: w ═ argmaxP (S | T).
Grading according to the syntactic rules that the word vector is classified and matched with the maximum probability from high to low:
classification Rules
First stage Cam shaft
Second stage Position of
Three-stage Timing of
(4) Decision tree analysis
And establishing a decision tree aiming at the result of cellular word grading according to the syntactic rule.
Classification model 1
As shown in FIG. 2, the camshaft is in one stage, the position, switching and regulation are in two stages, and the sensor, the actuator, the response, the timing and the exhaust are in three stages. The corresponding solutions are respectively as follows: the sensor corresponds to a camshaft position sensor, an engine wire harness and an engine control unit, the actuator corresponds to the engine wire harness, the engine control unit and a camshaft adjuster, and the timing corresponds to the engine wire harness, the engine control unit, the camshaft adjuster, a camshaft tensioner and a timing chain, and so on. The solution pointed by the thick arrow indicates that the probability of replacing the fitting is much higher than that pointed by other thin arrows, and the probability is preset in the decision tree model according to a large amount of maintenance data. Therefore, through carrying out word vector decomposition on the fault description of camshaft position (cylinder row 2) -timing over-advance and searching in a cell word bank, three cell words of the camshaft, the position and the timing are obtained, and through calculating the weight of the cell words and arranging the cell words according to a syntactic rule, the camshaft is located at the first level, the position is located at the second level and the timing is located at the third level. Through decision classification of the decision tree model, the most likely maintenance solution is to replace or maintain the camshaft tensioner.
Classification model 2
As shown in fig. 3, the word vector is decomposed for the failure description sentence "camshaft position sensor ═ sensor unreliable signal" and searched in the cell word bank, and three cell words of camshaft, position, and sensor are obtained. And calculating the weight of the cell words according to the three cell words, and arranging the cell words according to a syntactic rule to obtain that the cam shaft is positioned at the first level, the position is positioned at the second level, and the sensor is positioned at the third level. And then, the measured value of the component is combined to judge the accessory needing to be replaced. First, whether the wire plug has poor contact or not and the needle head is bent are checked. If so, the wire plug needs to be replaced. If not, entering a next-level decision tree and checking whether the voltage supply of the camshaft sensor is between 4.5 and 5.5V. If so, the signal voltage of the engine control unit is checked to see if it is between 4.5-5.5V. If otherwise the engine wiring harness needs to be replaced. By analogy, a maintenance solution is given by combining decision classification of the decision tree model with component measurement values.

Claims (2)

1. A fault remote diagnosis platform for vehicle faults, comprising:
the fault information receiving unit is used for receiving a fault description statement T; for a fault description statement T, its probability calculation formula is:
P(T) ＝ P(w1,w2,w3,…,wn)
＝ P(w1)×P(w2|w1)×P(w3|w1,w2)×…×P(wn|w1,w2,…,wn−1)
≈ P(w1)×P(w2|w1)×P(w3|w2)…P(wn|wn-1)
the fault information analysis unit is used for cutting words in the cell word bank to search cell words and arranging the cell words;
the fault information solving unit is used for providing a maintenance solution through decision classification of the decision tree model; the fault information analysis unit comprises a fault description statement decomposition unit, a retrieval unit, a calculation unit and an arrangement unit;
the fault description statement decomposition unit is used for cutting words of the fault description statement T in the cell word stock;
the retrieval unit is used for retrieving in the cell word bank;
the calculating unit is used for calculating the weight of each cell word;
the arrangement unit is used for arranging the cell words according to a syntactic rule;
the retrieval unit carries out word segmentation retrieval;
the method for calculating the weight of each cell word by the calculating unit comprises the following steps:
s1, calculating chi-square statistic of each cell word;
s2, taking the ith cell word with the highest chi-square statistic score, and calculating the occurrence frequency of the ith cell word in the jth fault description sentence T;
the calculation method of the chi-square statistic score in the step S1 is as follows:
weight=round(10×(1+lg(tfij))/(1+lg(lj)))；
wherein tf isijFor the number of times the ith cell word appears in the jth fault description statement T, ljThe length of the jth fault description statement T;
diagnosis method of fault remote diagnosis platform in automobile fault
(1) Constructing a base corpus
The method of creating the base corpus is to give a string whose natural language probability is P (w)1,w2,w3,…,wn),w1To wnThe cell words of the sentence are in turn;
the model is based on the assumption that the occurrence of the nth word is only related to the first n-1 words and is not related to any other word, and the probability of the whole sentence is the product of the occurrence probabilities of the participles;
for a fault description statement T, its probability calculation formula is:
P(T) ＝ P(w1,w2,w3,…,wn)
＝ P(w1)×P(w2|w1)×P(w3|w1,w2)×…×P(wn|w1,w2,…,wn−1)
≈ P(w1)×P(w2|w1)×P(w3|w2)…P(wn|wn-1)；
(2) and (3) decomposing the word vector: dividing the word vector into a plurality of cell words by a fault description sentence decomposition unit; each fault description statement T is represented as an n-dimensional word vector, T = (w)1,w2,...,wn) Wherein w is1To wnFor each cell word contained in each word vector；
The fault description statement T is:
P0021A camshaft position (Cylinder Row 2) -over-timing Advance
The fault description statement T is represented by a word vector, which may be an n-dimensional word vector (w)1,w2,w3,…,wn) Represents; wherein, w1Being a camshaft, w2Is a position, w3Is a cylinder, w4Is a positive timing, w5Is excessive; selecting each cell word as a characteristic item, wherein wiRepresenting the ith characteristic item;
calculating chi-square statistic of each cell word, and taking the ith cell word with highest score in the training set as a representative word to find out the word vector (w)1,w2,w3,…,wn) The least relevant cell word;
calculating, by a calculation unit, a weight of each cell word;
withe formula for calculating the weight of (a) is:
weight=round((10×(1+lg(tfij))/(1+lg(lj))),3)；
wherein tf isijFor the number of times the ith cell word appears in the jth fault description statement T, ljThe length of the jth fault description statement T; rounding the Round function to a specified decimal place, which is the third place after the decimal point in the above formula;
the feature expression of the fault description statement T word vector after the weight is calculated is as follows:
vector participle Weight of Cam shaft 0.286 Position of 0.191 Cylinder 0.048 Timing of 0.048 Transition of 0.048
(3) Search arrangement cell word:
after a user inputs a fault description statement T, a fault description statement decomposition unit in a fault information analysis unit is used for carrying out word vector decomposition on the fault description statement T so as to divide the complete statement into a plurality of cell word combinations; for each cell word, performing word segmentation retrieval in a cell word bank by using a retrieval unit in the fault information analysis unit; if the cell words are searched, calculating the weight of the cell words by using a calculating unit in the fault information analyzing unit; if the cell words are not searched, storing the cell words which are not searched in the time into a newly-added industry cell word bank, searching the rest cell words, and calculating the probability of the cell words;
calculating weights according to all cell words retrieved from a cell word bank, and arranging the cell words from large to small according to the weights by using an arrangement unit in a fault information analysis unit;
(4) and (3) analysis results:
the fault description statement T is:
P0021A camshaft position (Cylinder Row 2) -over-timing Advance
Each cell word retrieved from the cell word bank is weighted by a computing unit in the failure information analyzing unit, and the cell words are arranged by an arranging unit in the failure information analyzing unit according to the weights from large to small:
characteristic item Weight of Cam shaft 0.286 Position of 0.191 Actuator device 0.095 Sensor with a sensor element 0.048 Conversion 0.048 Response to 0.048 Transition of 0.048 Circuit arrangement 0.048 Cylinder 0.048 Regulating 0.048 Exhaust of gases 0.048 Timing of 0.048
Using discriminant analysis:
the word vector sequence of the fault description statement T is T = (w)1,w2,w3,...,wn) The sequence of each cell word in the fault description statement T after the weight calculation is S = (S)1,s2,s3,...,sn) (ii) a The calculation formula of the word vector classification matching maximum probability is as follows: w = argmaxP (st | T);
grading according to the syntactic rules that the word vector is classified and matched with the maximum probability from high to low:
classification Rules First stage Cam shaft Second stage Position of Three-stage Timing of
(5) Decision tree analysis
And establishing a decision tree by a fault information solving unit according to the result of the cell word grading by the syntactic rule.
2. The fault remote diagnosis platform for the automobile fault according to claim 1, characterized in that: the fault information resolution unit provides a maintenance solution in combination with the component measurement values.
CN201610364696.5A 2016-05-27 2016-05-27 Maintenance decision tree/word vector-based fault remote diagnosis platform Active CN106054857B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610364696.5A CN106054857B (en) 2016-05-27 2016-05-27 Maintenance decision tree/word vector-based fault remote diagnosis platform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610364696.5A CN106054857B (en) 2016-05-27 2016-05-27 Maintenance decision tree/word vector-based fault remote diagnosis platform

Publications (2)

Publication Number Publication Date
CN106054857A CN106054857A (en) 2016-10-26
CN106054857B true CN106054857B (en) 2019-12-24

Family

ID=57175981

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610364696.5A Active CN106054857B (en) 2016-05-27 2016-05-27 Maintenance decision tree/word vector-based fault remote diagnosis platform

Country Status (1)

Country Link
CN (1) CN106054857B (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107832908A (en) * 2017-09-29 2018-03-23 深圳供电局有限公司 A kind of standby redundancy needing forecasting method based on defective data
CN108154244A (en) * 2018-01-24 2018-06-12 南京天溯自动化控制系统有限公司 The O＆M methods, devices and systems of real estate power equipment
CN117590837B (en) * 2024-01-18 2024-03-29 深圳市伟创高科电子有限公司 Electric vehicle controller fault diagnosis method based on tree structure

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102200487A (en) * 2010-03-24 2011-09-28 通用汽车环球科技运作有限责任公司 Event-driven fault diagnosis framework for automotive systems
CN103106275A (en) * 2013-02-08 2013-05-15 西北工业大学 Text classification character screening method based on character distribution information

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102521613B (en) * 2011-12-17 2014-01-08 山东省科学院自动化研究所 Method for fault diagnosis of automobile electronic system
CN105159998A (en) * 2015-09-08 2015-12-16 海南大学 Keyword calculation method based on document clustering

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102200487A (en) * 2010-03-24 2011-09-28 通用汽车环球科技运作有限责任公司 Event-driven fault diagnosis framework for automotive systems
CN103106275A (en) * 2013-02-08 2013-05-15 西北工业大学 Text classification character screening method based on character distribution information

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于文本分类和SVM 的雷达侦察装备故障诊断研究;仝奇;《电光与控制》;20160228;第94-98页 *

Also Published As

Publication number Publication date
CN106054857A (en) 2016-10-26

Similar Documents

Publication Publication Date Title
CN111309912B (en) Text classification method, apparatus, computer device and storage medium
CN106095785B (en) Fault code diagnosis vehicle work item and spare part retrieval method based on decision tree classification
CN106021545B (en) Method for vehicle remote diagnosis and spare part retrieval
CN106054857B (en) Maintenance decision tree/word vector-based fault remote diagnosis platform
US10460240B2 (en) Apparatus and method for tag mapping with industrial machines
JP4860903B2 (en) How to automatically index documents
CN112597038B (en) Software defect prediction method and system
CN111427775B (en) Method level defect positioning method based on Bert model
CN112508105B (en) Fault detection and retrieval method for oil extraction machine
CN106054858A (en) Decision tree classification and fault code classification-based vehicle remote diagnosis and spare part retrieval method
CN111177655B (en) Data processing method and device and electronic equipment
CN111399464A (en) Intelligent vehicle quality inspection method and system
CN106649557B (en) Semantic association mining method for defect report and mail list
CN101276358A (en) Patent association analysis method
CN106844482A (en) A kind of retrieval information matching method and device based on search engine
CN117290404A (en) Method and system for rapidly searching and practical main distribution network fault processing method
CN111104422B (en) Training method, device, equipment and storage medium of data recommendation model
CN112015779A (en) Method, system and device for predicting preference of students
CN115577694B (en) Intelligent recommendation method for standard writing
CN111507534A (en) Student learning data-based predictive analysis algorithm for knowledge point mastering conditions
CN106056223B (en) Platform for vehicle remote diagnosis and spare part retrieval
KR20200094514A (en) Apparatus and Method for Extracting Facility Error Pattern
CN112559798B (en) Method and device for detecting quality of audio content
CN106055439B (en) Based on maintenance decision tree/term vector Remote Fault Diagnosis system and method
CN114820618A (en) Defect detection model training method, device, equipment and storage medium

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant