The faintness in breakdown diagnostician system pushs manage algorithm 

  • Time:
  • Click:336
  • source:WEDEMEYER CNC Machining
Summary purpose studies the ambiguous inference of breakdown diagnostician system is algorithmic. A variety of circumstances in will normal inference and conjecture inference handle the method apart, use retrorse symbol inference and the method that are united in wedlock to digital inference photograph, use ambiguous theory to design evidence to input interface. The result put forward a kind of breakdown to diagnose algorithm of ambiguous and integrated inference, developed system of FMS breakdown diagnostician. Knot logic talks analysis and application to make clear, this algorithm diagnoses a domain to have versatility and efficient sex in breakdown. Keyword breakdown is diagnosed; Ambiguous and retrorse inference; to digital inference; Evidence inputs interface to classify date TP18; Wang Xinyi(School Of Mechanical Engineering And Automation of   of   of Shi Tianyun of   of   of TP277The Synthetic Fuzzy Inference Algorithm In Fault Diagnosis Expert SystemYuan Hongfang, beijing Institute Of Technology, to Study The Synthetic Fuzzy Algorithm Applied In Fault Diagnosis Expert System of   of Aim of   of Beijing   100081)Abstract.

Methods   According To Many Cases In The Normal Inference And Conjecture Inference, the Method Combining The Back Symbol Inference And Forward Digital Inference Was Adopted.

The Input Interface Of Evidence Was Designed Using The Fuzzy Theory.

Results   A Synthetic Fuzzy Inference Algorithm For Fault Diagnosis Expert System Was Advanced, and The FMS Fault Diagnosis Expert System Was Developed.

Conclusion   By The Analysis Of Theory And Application, the Fault Diagnosis Algorithm Has The Characteristic Of Universality, completeness And High-efficiency In The Fault Diagnosis Field.

Breakdown of Fault Diagnosis; Fuzzy Back Symbol Inference; Fuzzy Forward Digital Inference; Input Interface Of Evidence1 of Key Words   diagnoses integrated inference to calculate law document [1] the algorithm of inference of ambiguous and retrorse symbol of the consideration sense that offer, intuitionistic phenomenon and the nerve network that are based on conjecture to inferential algorithm, better land realized a variety of fast and effective inference. But consider a law medium in retrorse symbol inference not will individual be diagnosed continuously and diagnose departure, also not will exterior be in with in-house diagnostic knowledge in inference distinguish clearly orderly early or late, did not consider the conjecture inference issue of equipment subsystem in conjecture inference algorithm. The author goes up in the base that improves these blemish, the faintness of breakdown diagnostician system that put forward to have common sense pushs adjustment way integratedly [2] , following plan institute show its flow chart. Breakdown diagnoses integrated inference to consider system of law flow chart self-recording in inferential process frequency of every regular diagnostic success and diagnose total number, the assimilate to of both uses successful rate for regulation, bigger to using power regular self-correcting is corresponding degree of confidence, in knowledge base system, user of lesser to regular and successful rate regular and active reminder undertakes be revised accordingly or be deletinged, the ego that realizes knowledge is perfect. The basic idea with algorithmic algorithm of inference of 2 ambiguous and retrorse symbols is: Set out from breakdown phenomenon or target of reason barrier module, adopt particular search strategy and strategy of conflict clear up, all premise of may contented target are found out in the knowledge base, quiz to the user according to corresponding knowledge, undertake the answer of the user and knowledge base premise blurring matching, deliver algorithm according to degree of confidence, the degree of confidence of computational target. When satisfy, continue to be down a search, otherwise backdate, relapse so until find out first breakdown account. After the system investigates breakdown matter, inferential machine chooses according to the user, continue to search other breakdown account, be eliminated till systematic breakdown or all over all previous whole knowledge base. Algorithm of breakdown diagnostic inference searchs strategy to use heuristic deepness commonly preferential search method and backdate method [3] . Strategy of clear up of the conflict in this algorithm is adopted from equipment exterior the optimum seeking method that is united in wedlock to in-house diagnostic regulation and regular degree of confidence, when namely should many regulation matchs, by equipment exterior diagnose regulation to interior, degree of confidence from arrive greatly small order is ordinal choose. When regular degree of confidence is equal, inferential machine will deposit in regular library according to its early or late order is ordinal choose. But when should having clear sense, intuitionistic appearance, the regular and first grade of existence sense organ and intuitionistic phenomenon is highest, it is next from equipment exterior diagnose regulation and regular degree of confidence to interior from arrive to choose a principle smally greatly. Of the premise condition of ambiguous regulation and fact match rate by match function decision, the product that the degree of confidence of conclusion matchs value and regular degree of confidence for this [1] . The biggest advantage that 3 faintness are calculating law number deduction to digital inference is inference in need not regular premise and conclusion match stage by stage [4] , undertake the number is calculated directly however, rate is accordingly rapidder. Breakdown introduces digital inference in diagnostic inference algorithm is to decrease mutiple level match, improve inferential efficiency. Faintness is pushing adjustment way to the number is in retrorse inference algorithm ambiguous regulation matchs the base that reachs degree of confidence to transmit a method to go up, omit ambiguous and regular search is reached mutiple level match, condense algorithm entirely to learn formula. Formulary calculation is used directly in inference, can gain conclusion degree of confidence quickly. To " with " ambiguous generation pattern is regular: IF A1(b1) And A2(b2) And... And An(bn) THEN Q(u) , among them Ai is premise condition; Bi is the degree of confidence of premise, i=1, 2, ... , n; Q is conclusion; U is regular degree of confidence. The initiative evidence that obtains suppose is A1(b1) , a2(b2) , ... , an(bn) , among them Ai is initiative evidence, bi is evidential degree of confidence, criterion the degree of confidence of conclusion is E and ∧ of =(t1 ∧ T2... ∧ Tn)u namely E and =min(t1, t2, ... , tn)u, among them T1=1-max{0, b1-b1};t2=1-max{0, b2-b2}; ... ;tn=1-max{0, bn-bn} . To " or " ambiguous generation pattern is regular: IF A1(b1) THEN Q(u1) ... IF An(bn) THEN Q(un) , criterion the degree of confidence of conclusion is E or =1- [(1-t1u1)(1-t2u2) , ... , (1-tnun) ] , among them T1, t2, ... , tn is worth Alexandrine type. When be being used actually, because library of every breakdown phenomenon is medium premise amount, with or regular relation and knowledge base number of plies are different, can decide respective computational function or the means with formal parameter call general and algorithmic function to come true respectively. To doing not have the premise of the choice to the user in conjecture inference, algorithm should acquiesce automatically its degree of confidence is 0. 4 evidence input interface to design breakdown to diagnose the origin of initiative evidence to have two kinds: The evidential knowledge that ① forms according to the answer of the user; The evidential knowledge that ② forms by the answer of the user sets out, use the intermediate result of generation of ambiguous inference algorithm. In inferential process, ① by inferential machine basis in the knowledge base additional inquiry raise a question to the user, in order to decide the degree of confidence of this premise; ② has consideration automatically by inferential machine. Evidence inputs interface to design be in the light of ① character. According to knowledge base regulation, the problem may have 3 kinds: If ① is called regulation is determinism (regular to ambiguous generation type, if regular premise degree of confidence is 0, think this regulation is determinism namely, consider as the) of uncertainty otherwise, the question that raises will be affirmatory, the user wants the answer that gives out to decide or deny only, inferential machine will call regular evidential degree of confidence according to replying 1 or - 1. If ② is called regulation is uncertainty, the question that raises will be ambiguous, the user can input degree of confidence directly, also can choose from inside ambiguous measure word, the system expresses the basis quantized interval of 1 medium ambiguous measure word, the faintness that quantifies an user replies, gain evidential degree of confidence. Those who express ambiguous measure word quantify ambiguous measure word to be sure special may probable likelihood is a bit likely possibility is lesser possibility is little cannot numeric interval [1.

00, 1.

00] [0.

93, 0.

99] [0.

80, 0.

92] [0.

65, 0.

79] [0.

45, 0.

64] [0.

30, 0.

45] [0.

01, 0.

29] [0, 0] the 1 load that set IF crosses exemple of <DIV Align=left>   big (Subject=0.

80) THEN main shaft communicates CF=0 of electric machinery overheat.

80 among them Subject=0.

80 express premise degree of confidence, CF=0.

80 express regular degree of confidence. User but the experience according to oneself, give out directly the degree of confidence with too big load. If the user forbids to inputting degree of confidence to hold, also can choose ambiguous measure word. The ambiguous measure word that chooses suppose is " likelihood " , the correspondence that expresses on the basis concerns, "Likelihood " for [0.

65, 0.

79] , take median 0.

72, so degree of confidence of laden too old evidence is 0.

72, and premise degree of confidence is 0.

80. Of regular premise and evidence match degree to be 1-max(0, 0.



92. Set regular premise and evidence to match threshold value for 0.

90, because of 0.

92 > 0.

90, so regular premise and evidence match a success. The degree of confidence of conclusion is (1-0.

08) × 0.

80 ≈ 0.

74. The threshold value that establishs conclusion to hold water is 0.

60, as a result of 0.

74 > 0.

60, criterion this breakdown happens. If ③ is called include in regulation can detect quantity, at the same time this regulation is not affirmatory. Criterion inferential machine measures requirement user a few check on the spot, give out next numeric. Systematic basis this are numeric, transfer regular subject function, the degree of confidence of computational evidence. About the choice of subject function, decide by experience and experiment. Exemple 2 set one regulation to be pressed too for IF voltage (Subject=0.

70, formula=5, scope=(220, 210, work of 230)) THEN system is abnormal CF=0.

70. Among them Subject=0.

70 express premise degree of confidence, formula=5 expresses to use the 5th kind of subject function, rise half times namely model distributinging function </DIV>scope=(220, 210, 230) states the normal value of voltage is 220V, of voltage liminal for (210V, 230V) . In diagnosing a process suppose, detect the real value of voltage is 225V, criterion the degree of confidence that voltage presses too is B= [(225-220)/(230-220) ] 2=0.

25, among them A=220, b=230, x=225. By can calculating directly to digital inference algorithm the degree of confidence with a systematic abnormal work is E= [1-max(0, 0.


25) ] × 0.

70 ≈ 0.

32 < 0.

60, because this premise won'ts do,stand. The system of FMS breakdown diagnostician that 5 conclusion develop with this integrated inference algorithm already center of test of try out Yu Changchun FMS, preliminary moving result makes clear: This are algorithmic not only inferential rate is rapid, ambiguous inference is reasonable and effective, and consideration element is comprehensive, inference reliable, efficiency is tall. Because designed a variety of user evidence to input interface, avoided a tradition to input the difficulty of degree of confidence. This inference algorithm is compositive at the same time retrorse symbol inference and to digital inference, omnibus tall, specific aim is strong, agree with the need of breakdown diagnostic domain. CNC Milling