论文研究-基于EMICA-KRR的长输管道压力监测与泄漏定位方法.pdf-资料库

weixin_38744270-11795946-4744302543332527631.pdf-第1页.png

第1页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第2页.png

第2页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第3页.png

第3页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第4页.png

第4页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第5页.png

第5页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第6页.png

第6页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第7页.png

第7页 / 共11页

weixin_38744270-11795946-4744302543332527631.pdf-第8页.png

第8页 / 共11页

39 7 2019 7 Systems Engineering — Theory & Practice Vol.39, No.7 July, 2019 doi: 10.12011/1000-6788-2017-2207-11 : X913.4 : A EMICA-KRR , ( , 710055) , "$%&’, )* +,-0 2 3 (EMICA) 45678 (KRR) %9$. , 0* EMICA %9@, A BD%EG4EG I JK2, 9G B 2; MN, OD EMICA @PQ%9G, RI * KRR %9@, TUDQ)9GVWX, G% 9%$; N TE (YZ[ - \) % !!"%#^. ‘ab: EMICA-KRR $%cde%G B^, f"h9G "$ $(, klmnK%oÆpqr. 9; ; 0 2 3; 5678; $; YZ[ - \ A long distance pipeline pressure monitoring and leakage location method based on EMICA-KRR ZHANG Xinsheng, WANG Zhe (School of Management, Xi’an University of Architecture and Technology, Xi’an 710055, China) Abstract In order to deal with the problem of sudden leakage during pressure monitoring of long-distance pipelines and being diﬃcult to give early warning and accurate leakage location, a pipeline leakage fault detection and location model is proposed by using ensemble modiﬁed independent component analysis algorithm (EMICA) and kernel ridge regression algorithm (KRR). Firstly, a fault detection model based on EMICA algorithm is established, which extracts and separates Gaussian and non-Gaussian signals from pressure data and construct related statistics to achieve fault signal separation and principal component selection. Then, based on the fault signals obtained by the EMICA model, a fault diagnosis model by using the KRR algorithm is further constructed, and the data is ﬁtted to obtain the amplitude of the pressure change of the fault signal, and the leakage signal selection and the location of the leakage fault are achieved. Finally, numerical simulation experiments of the TE (Tennessee-Eastman) process were performed to verify the performance of the proposed algorithm. The simulation results show that the EMICA-KRR algorithm has better signal separation capability, and can accurately identify the leakage fault signal and accurately locate the failure position of the pipe segment, which overcomes the shortcomings of the traditional methods such as ineﬃciency and delay. Keywords fault diagnosis; pressure monitoring; independent component analysis; kernel ridge regression; leak location; Tennessee-Eastman process : 2017-12-14 s : (1978–), , , , , , : , E-mail: xin- sheng.zh@outlook.com; (1993–), , , , , : , E- mail: xauatWZ@163.com. : (41877527); (2016JM6023) Foundation item: The General Program of National Natural Science Foundation of China (41877527); Natural Science Foundation of Shaanxi Province of China (2016JM6023) Æ !": , . EMICA-KRR [J]. , 2019, 39(7): 1885–1895. !": Zhang X S, Wang Z. A long distance pipeline pressure monitoring and leakage location method based on EMICA-KRR[J]. Systems Engineering — Theory & Practice, 2019, 39(7): 1885–1895.

1886 1 #$ 39 , , , . , , . !"$. ! %, &# $!!"#, "’($ % $. ’&’), !%, . ! , (-# !"# !$’ [1−4], $#$(.)!)*0. 2, *3+%% &, & , Æ$ 4 . -% [5] ’&$)()*)(’ (+ *() ; )* [6] ) *(6() *-0$1 ),)* , + +# ,&-; -* [7] ’3. (/ -#Æ, .9 -#, :’$.’&;; . [8] &$)()*)(")) *) , ’ Lasso 01 /*(, //*() /0 , +! Lasso 01<+")!%,; Tong[9] &$)()* 3, 03451)-&$)(1, +!&26, , ?&’. 4, 27!7,3’. 3+@)*, Æ8Æ3+"5&$)()*< (ensemble modiﬁed independent compo- nent analysis, EMICA) 401< (kernel ridge regression, KRR) ). +’&’), ,66, Æ8#91!7, !9$" T 2 Q ( , 0$"58<))(); 8, ’)!9 , 0$401< 01)*, 9/ , 0$*! ’!9 X ∈ Rm×n, n >Æ, m *(. X : d ;&$ ’=(9. , d ≥ m (@ d = m, >9;9 ). &$)()*0$>)9 W @ S ;(; A, !A9<6: X = AS + E. ˆS = W X. (2) Æ<’0, B=)(, :=ÆC’/)(, ?*-

7 &", Q = Λ−1/2P T, -8 Z :@")@ . , : EMICA-KRR Æ)(D&$, (><6 CTC = D E/A9 C ∈ Rm×n, ?!")@ 1887 S, ,FHE@: &", D = diag {λ1, λ2,··· , λm}. !<6 SST/ (n − 1) = D E8, ! S "?&$)(>B, )(’IG, !91 S = CTZ. (5) -8 Sn, E@D: Sn = D −1/2S = D −1/2CTZ = Cn TZ. 0$>@/A9 Cn (&" Cn <6 Cn Rd×m ;( A ∈ Rm×d, &FHE@: T = D −1/2CT Cn (6) TCn = I), :!)9 W ∈ −1/2Cn −1/2Cn −1/2P TA = P Λ −1/2CnD −1/2. W = D (7) &", W A = Id ∈ Rd×d. K!$@L<8, F=A#G$A >*- T 2 Q ( , &FHE@: TQ = D TΛ &", s = W x, e = x − As. 2.2 012345 T 2 = sTD −1sQ = eTe. (8) +Æ!9)(, (M#&!H &$. ?"II;, ’:,(Æ)( &$’. , BB,()(’: J (y) = [E {G (y)} − E {G (v)}]2 , &", y ;2BE >, v !;2BE >*(, G ! K. 0$1/A K, :!9/A B. Hyv¨arinen [11] ÆO K: G1 (u) = G2 (u) = exp 1 2 log cos h (a1u) , −a2u2/2 , (9) (10) (11) (12) &", 1 a1 ≤ 2, a2 ≈ 2. ! K1 , 0$!)7, :8: 7)L/A K, ?!9.’/)K). 2.3 y7 49 G3 = u4. )!9)(0$PM6, :&")(. ?+)(:# (CPV) 1"))*Q>7)(@ )(. 2) EMICA-L2: 9?O)(. 3) EMICA-nG: !’IG, )(’IG&)(@)(. @L6"R+$, !3+ [13] "1 . ?+ $%, !?91;-6@, (PM.’1. +PM "A6)&’, Æ8"O KO)(’PMH, EJ 1 "D. 0$!745KA!=(HI, 0$T)L:9P M26, A)$")(1,.

1888 39 2.4 * .4:; 19#$"7!CKNO, Q(’ K1$AAJ)( 1. &’ !?91;-6, / K)(PM26;(F" 3"PL [14]. "58)$E1 K, !/O:0’"1(, 5)E @D: j i + E j i , j X = A i S j j i = W i X. S (13) &", i ∈ {1, 2, 3}, j ∈ {1, 2, 3}. "58 9 1, K)(PM26:0=(MV&". EJ 1 D, C (, ’45K/&RH=("I [15]. X PCAZ=QX X = S W x j j i i T and Q 2 i, j i,j BIC 2 T BIC Q 1G 2G 3G MICA-CPV MICA-L 2 MICA-nG { j j j X=A S E + i i i S W X j j i i = with j { } 1,2,3 and i { } 1,2,3 3 3 =∑∑ i 1 j 1 = = 3 3 =∑∑ i 1 j 1 = = 2 i P (x | F) P (F | x) 2 T T i 3 3 P (x | F) ∑ ∑ 2 T i 1 j 1 = = i P (x | F) P (F | x) Q Q 3 3 P (x | F) ∑ ∑ Q i 1 j 1 = = > 1 ?B (EMICA) DEI> i (F ) , PT 2 i (x|F ) PT 2 i (x) i (N) + PT 2 PT 2 i (F ) . i (x|f) PT 2 PT 2 i (x) = PT 2 PT 2 (F|x) = i (x|N) PT 2 − T 2 3 3 T i,lim 2 i (14) (15) &", N F )#

7 , : EMICA-KRR 1889 +’$", +’ S-S’#-’), Æ81401< @O)3. / 0/3+>Æ, KRR .’.%, : , u 3QQ%,, v .’% , ([3VTQVR’TY’U). 5 { 1Æ8Æ #’, Æ ! TE $ [19] @BZ V1. TE $!"#P@$3], 0[(=\ 52 *($, :Æ 21 7/6. &%) #VBÆ, .(’$ ZV. 1 EMICA )#’, B9$ [20] ZV:, )*3+O K G1-MICAG2-MICAG3-MICA EMICA !R )#, E@*(:

1890 ⎡ ⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎣ 0.95 0.82 0.94 0.23 0.45 0.92 0.61 0.62 0.41 0.49 0.79 0.89 0.89 0.92 0.06 0.76 0.74 0.35 0.46 0.18 0.81 0.02 0.41 0.01 39 s + . (22) ⎤ ⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎥⎦ x = As = &", A ;(9, s *(, ( <6;2B 0.01 >*(. "*(!E@ D: s1 (k) = 2 cos (0.08k) sin (0.06k) , s2 (k) = sin (0.3k) + 2 cos (0.1k) , s3 (k) = (rem (k, 27) − 13) /9. (23) (24) (25) &", k E, rem(k, 27) Æ0 k 27 8 . 1 XZQÆÆ[\]^ ZW S^ 1 2 3 100 Y[], T A (5, 2) _] 0 100 Y[], Z‘]U s2 V]^ 100 Y[], ÆV]^[_^ x1 > 2 P|QÆÆBST> *( si (i = 1, 2, 3) !/E@" 1000 >Æ, EJ 2 D. ?+*(!?*( , :&’!$"#. $", O)(, 7 ] 1 − α = 99%. !TR, O/E< 1 D. ?+76@37! , 1-#$//‘ I\a@*- )W, F& -#0X(;[LU. ?+’ ! !, )+ (Y,)S -,. ! @, !9 G1-MICAG2-MICAG3-MICA EMICA )]Y, EJ 3∼6 D. > 3 G1-MICA _‘ad

7 , : EMICA-KRR 1891 > 4 G2-MICA _‘ad > 5 G3-MICA _‘ad > 6 EMICA _‘ad A]Y:Z, EMICA )&4&O K) )#. &", G1- MICA EMICA / 1 -#,&, d: G2-MICAG3-MICA ! BICQ (@:!9 N/, +&#:&-. :J 5 J 6 ]Y:Z, G3-MICA =00$ ( BICQ 9 3, G1-MICAG2-MICA _=00$ BICT 2 ! 3 7!. ::Z, EMICA + 3 :7 0$ BICQ BICT 2 !. @, :ZV)L:!9, KE 10. @. EMICA ):0$3451K/3)=

1892 39 ("Q, []"PL , Æ )#S-0. 6#L:, Æ8! TE 3$@Æ 21 "7/6@ G1-MICA, G2-MICA, G3-MICA EMICA ):, !9c)LE< 2 D. d S^ A/C \]e, B ^fg B ^f, A/C e^g D \]‘h aÆf\‘h hf\‘h i] C b A,B,C i]^f D \]‘h C \]‘h 1 2 3 4 5 6 7 8 9 10 11 aÆf\‘‘h 12 hfk‘‘‘h 13 aÆi 14 aÆfkb 15 hfkb 16 _ 17 _ 18 _ 19 _ 20 _ 21 adbbc‘ 4 2 TE QÆÆ[^Befg (%) EMICA G1-MICA G2-MICA G3-MICA BICT 2 BICQ BICT 2 BICQ BICT 2 BICQ BICT 2 BICQ 0.25 1.72 97.6 0.00 1.12 0.00 0.00 2.1 2.8 11.6 25.2 1.50 4.5 0.00 5.4 9.5 3.71 10.0 12.6 9.5 46.7 0.13 1.58 91.5 0.00 2.50 0.00 0.00 2.0 1.2 15.2 29.0 2.43 5.1 0.00 12 13.4 6.0 10.0 22.4 18.1 44.0 0.25 1.88 98.25 0.00 0.12 0.00 0.00 0.21 1.9 13.2 28.6 3.2 4.9 0.00 3.6 11.2 4.6 10.2 15.0 11.0 47.0 0.13 2.00 93.00 0.00 0.53 0.00 0.00 0.19 6.5 16.5 31.2 2.5 5.00 0.00 12.9 13.7 6.4 9.5 24.2 19.2 44.6 0.25 1.63 97.50 0.00 0.28 0.00 0.00 0.24 1.8 13.7 44.1 1.2 4.9 0.00 4.6 11.9 8.1 9.9 18.7 10.5 48.7 0.13 2.38 95.63 0.00 0.17 0.00 0.00 0.19 3.3 16.1 30.1 0.11 5.1 0.00 9.9 17.0 6.7 10.2 29.9 17.9 47.2 0.13 2.25 98.63 0.00 0.33 0.00 0.00 0.19 1.5 12.9 29.7 1.32 4.7 0.00 4.9 11.4 3.9 10.0 25.4 9.6 53.4 0.50 1.88 97.25 0.00 0.26 0.00 0.00 0.21 5 18.00 32.5 3.6 5.4 0.00 6.3 17.1 7.0 9.9 26.2 19.9 44.9 < 2 "jQ!! TE $":. Q EMICA )L3, ’Æ8 KRR[21,22] O. < 3 CD KRR < : SVR[23] PCA +601)*)L. ; SVR PCA :, /& .’f5, 7 /01 <)L-

资料库

论文研究-基于EMICA-KRR的长输管道压力监测与泄漏定位方法.pdf

相关推荐

开发技术

热门标签

最新资料