KHOA HOC CONG NGHE
S60b'2026
M6 phong cuang dp be tong 28 ngay tudi SLT dung mang no-ron nhan tao toi uu hoa voi thuat toan giai thuat di truyen
• TS. LV H A I B A N G ; T S . N G U Y E N T H O Y ANH TrUdng Dgi hpc Cong nghe Giao thdng van tdi
T 6 M T A T : Mang ludi than kinh nhan tao gan day da duoc su dyng rdng rai de mo phdng nhieu bai toan trong ITnh vuc ky thugt xay dung. Trong nghien cuu nay, mo hinh mgng no-ron nhan tgo (ANN) v6i thuat toin Levenberg - Marquardt (LM) v i mo hinh ANN toi uu hoa b i n g giii thuit di truyen (ANN- GA) duoc xiy dung di dy doan cudng do nen cda bfi tong cudng do cao (HPC) 6 28 ngay tuoi. Hai md hinh tr§n dupe phat trien thong qua qua trinh huin tuyen v i kiem chung, sd dung 425 ket q u i th( nghtem tu cac nghien cuu quoc te Cac tham so d i u vio cua b i i t o i n la xi m3ng, xl 16 cao, tro bay, nuoc, phy gia sieu d^o, cot lieu tho, cot lieu mm v i tudi HPC, cudng do nen cua HPC la ham myc tieu cua m6 hinh mo phdng Hai tieu chi la he so tuong quan (R) va sai so t o l n phuong trung binh (RMSE) di duoc sd dyng de d i n h gia hieu q u i dy bao cic mo hinh de xuat Ket q u i cho thay, c i hai md hinh deu c6 khi nSng dy doan tot, trong do m6 hinh ANN-GA co hieu q u i dy bao cao han Nghi4n cuu nay cho thay cac mang than kinh nhan tao CO tiem nSng cao de dy doan gia tri cudng do HPC v i cho thIy tinh hieu q u i cua cong cy toi uu hda mang no-ron.
TL/ K H O A : Tri tue nhSn tao (Ai), mang no-ron nhan tao (ANN), be tong higu suit cao (HPC), cudng dp nen, giii thu|t di truyen
ABSTRACT: Artificial neural networks have been widely used to solve many problems in many areas related to civil engmeenng In this work, ANN with Levenberg-Marquardt algorithm and ANN optimized by Genetic Algonthm (ANN-GA) were developed to predict the compressive strength of high-performance concrete (HPC) at the age of 28 days. In order to construct the models, the training and testing datasets were developed using 425 experimental results gathered in the literature The input parameters of the problem were the contents of cement, blast furnace slag, fly ash, water, superplasticizer, coarse aggregate, fine aggregate, and testing age, whereas the compressive strength of HPC was the predicted
output Two commonly used quality assessment criteria, namely the correlation coefficient (R) and root mean square error (RMSE), were used to validate and compare the performance of the developed models The results showed that both models exhibited good predictability, but the ANN- GA was more effective than ANN The results of the study showed that ANN exhibited a high potential in predicting the compressive strength of HPC, and GA was successfully used to improve the prediction performance of ANN
KEYWORDS- Artificial Intelligence (AI), Artificial Neural Network (ANN), High performance concrete, compressive strength, genetic algonthm.
I.DATVAND£
Be tdng la mdt loai v i t l i l u daoe sd dung rdng rai trong eie k i t elu edng trinh bdi nhilu tfnh n i n g die biet eua nd. Cung vdi sU p h i t triln cua eie edng ngh^
ky thuit xiy dang, sd p h i t trien eua cdng n g h i b l tdng mdi cung ddng mdt vai trd rlt quan trpng. Ngiy cing cd nhilu loai v i t lieu mdi ra ddi v l da dapc dng dung rdng r i i trong cle k i t elu cdng trinh hien dai. 81 tdng cadng do cao (High Performance Concrete - HPC) l i mdt trong nhCmg loai v i t lieu mdi ed them cic tinh chit cO ly dapc c l i thien, mang lai s a t i l n bd trong cdng n g h i vat lieu va ket elu xiy dong. HPC dope hieu l i mdt loai be tdng xi mang vdi eic dac tinh eo b i n nha: cadng do chju n I n cao, chiu keo khi udn rat cao, md-dun d i n hoi cao, ben vdng v l dn dinh dadi eie t i c ddng x l m thac bat loi cua mdi trddng... va nhilu d i e tinh khic, rat hdu ich trong xiy dong ha t i n g giao thdng. Nhd sii cii thien eic tinh chat cO hpc n i y m i HPC dupe sCf dung rdng r i i trong eic cdng trinh xiy ddng, die b i l t l i trong cle tda n h i eao t i n g , xiy ddng dadng bd, elu dai v i dadng h i m [1 ]. Bin chat eua edng n g h i trong quy trinh sin xuit HPC l i l i m cho nd cang d i e c i n g tdt. Do dd, eie quy tie de p h i n biet HPC so vdi b l tdng thddng la: (i) Viee sd dung kich thdde cdt heu nhd, (ii) bd sung eie v i t lieu xi m i n g nhamadi silie hoic tro bay v l (iii) quan trong nhat l i dng dung cua phu gia sieu (hda) d i o d l g i l m ty le nadc tren chat k i t dinh [1]. De che tao HPC, cic vat lieu cau t h i n h
37
S6 062020
eung nhaty I I cua chung can p h l i dapc laa chpn v i kiem soit mdt eieh can than. Mdt sd nghiin edu trong cle t i i l i l u [2,3] da d l xuit eic phacmg p h i p t h i l t k l hon hpp cho HPC. Myc tieu chi'nh eua cic n g h i i n cOru n i y l i de cd dapc sa ket hop eie v i t l i l u c i u thanh v l ty le taong dng d l tao ra HPC vdi cle tinh chat eo hoc dape c i i thien.
Cadng dd n I n cua HPC I I mdt tfnh chit co hpc quan trpng, thu hut daoc sa quan t i m cua nhilu n h i nghiin edu. Hien nay, vile xic dinh eadng dp n I n eua HPC chu y l u doa t r l n viec thac hien eie thi nghiem trong phdng thf nghiem. Tuy nhiln, quy trinh thd nghiem n i y rat tdn thdi gian, tdn k i m v i edn yeu c i u mdt sd t h i l t bi die bilt. Do dd, cic n h i nghiin edu d i ed g i n g d l xuat mdt sd edng thdc chi ra mdi taong quan thac t l gida cadng dp n I n HPC vdl mdt sd tinh chat hoic thdng sd co hoc l i l n quan. Zhou v i edng sa [4] da nghien edU I n h hadng cua cdt lieu d i n cadng dd n I n HPC v l k i t luin ring nd cd t h i daoc da d o i n b i n g mdt so cdng thdc, ngoai trd cle tradng hpp cdt li§u cd md-dun r l t t h i p hole r l t cao.
Duval v i Kadri [5] da nghiin edu I n h hadng cua mudi silic d i n cudng dp n I n HPC v i dda ra md hinh da d o i n vdi he sd taong quan 0,991. Chan v l cdng sa [6] de xuit mdt md hinh lien quan d i n do ben v i dp xdp cua HPC sau khi t i l p xuc vdi nhiet do eao. Rashld v i cdng sa [7]
da nghien edu mdi taong quan cua mdt sd tinh c h i t co hpc, bao gdm md-dun d i n hdi, do b i n k l o v i I n h hadng eua kfeh thudc m l u vat den eadng dd nIn. Ndi ehung, eic nghien edu niy eung cap eie cdng thdc thae nghiem daa tren k i t q u i thf nghiem 64 xic dinh cadng dd nIn cua HPC mdt cich nhanh chdng nhang g i n dung. Tuy nhien, nhape diem eua cic phaong p h i p niy I I nd chi cd t h i tfnh t o l n daoc vdi mdt sd lupng tham sd gidi han trong xiy dang eie h i m quan h i . Do dd, eieh tiep cin phaong trinh thdc nghiem trd nen khdng thae te khi so laong tham sd d i u v i o Idn.
Gin diy, trong linh vac ky t h u i t xiy dang, cic phaong p h i p hien dai v i tien t i l n hon da dacK gidi thieu. trong dd cd the d l cap d i n ky t h u i t tn tue nhan tao (Artificial Intelligence - AI). Trong ky thuat xiy dang, tri t u l n h i n tao da daoc sd dyng v l p h i t trien cho muc dich d a d o i n v i tdi Uu hda khic nhau, ching han nha da d o l n cadng do n I n cua be tdng [8] v l d a d o i n eadng do e l t d i m be tdng gia ed spi [9]. Cic vi du khic l i xac dinh thiet hai trong elu true xdong [10] v l tdi au hda hinh dang eua d i p vdm [11], Sd dung mang no-ron n h i n tao (ANN) cd t h i cung cap g i l i p h i p l i m t i n g dp chinh xic cua vile tinh t o i n h i m muc tilu.Tuy nhien, eie md hinh khie nhau dua tren tri thdng minh n h i n tao cd the I n h hadng d i n h i l u suit cua viec tfnh t o i n [12].
Muc tieu chinh eiia nghiin edu n i y l i d l xuit cle md hinh AI phu hpp dua tren trf tue n h i n tao d l d a d o i n eadng dd n I n cda HPC. Vcri myc dich niy, md hinh ANN sd dung t h u i t t o i n Levenberg - Marquardt v i md hinh ANN dapc tdi au hda b i n g g i i i t h u i t di truyin (ANN-GA) ddoc p h i t tnen de d a d o i n v i so sinh hieu q u i . Md hinh de xuat da duoc thd nghiem vdi 425 dO lieu thf nghiem eddng dd n I n duoc thu t h i p t d cle edng bd qudc t l vdi 8
38
thdng sd d i u v i o v i eddng dd nen cda HPC 28 ngay tuoi I I ham muc t i l u .
2. CO s d DUf Li£u vA M O H I N H D U B A O 2.1. DCrligu thi nghiem
DO lieu nghien edu cua nghiin edU niy bao gdm 425 m l u HPC 28 ngiy tudi dupe thu thip td k i t q u i edng bd trong eie bii b i o [13-16]. Tit c l eie m l u dUOc c h i tao b i n g xi ming Portland thdng thadng (OPC) v i dope xd ly trong d i l u kiln binh thadng. Dii lilu thd nghiem HPC cd sin trong t l i lilu da sddung mau vit ed kich thdde v i hinh dang khie nhau. Chung dapc ehuyin ddi thinh ket C|ul cudng dd n I n vdi m l u hinh tnj 150 mm dua trln eie t i l u chuan hiln cd, nha IS 516 1959 va GB 50205 2001. Cadng dd n I n HPC daoc d a d o i n dua vio 8 biln d i u vao: xi mang (X,), xi Id eao (X^), tro bay (X^), node (XJ, phy gia silu d i o (Xj), cdt lieu thd (X,), cdt lieu min (X,) v i ngiy tuoi HPC (X^).
2.2. Mang no-ron nhin tao va thuit t o l n LevenJMrg Marquart
Trong nhdng t h i p ky qua. vile sd dung mang no- nan n h i n tao (ANN) da trd n I n phd biln trong ky t h u i t xiy dang edng trinh. Mang ladi t h i n kinh nO-ron n h i n tao l i mdt dang cua t h u i t t o i n hoe miy, md phdng b i i t o i n dua t r l n nguy6n ly sinh hoe cda bd n i o con ngadi.
Cie dieu kiln cua dd li^u d i u v i o v i cic nguyen ly eo hpc, v i t ly, hda hpc khdng c i n p h l i ddpc xic djnh rd r i n g khi sddung t h u i t t o l n hpc m i y ndi chung v i ANN ndi rieng.
Ky t h u i t md phdng dya t r l n ANN rlt hilu q u i trong vile tim ra g i i i p h i p cho cic v i n d l phdc tap m i eie md hinh t o l n hpc truyin thdng khdng g i l i quylt dupe.
Ktin true d i l n hinh cua m^ng ANN thUdng bao gdm ba lop: Idp d i u vio, Idp I n (bao gdm cie t l b i o than kinh) v i Idp d i u ra (Hinh 2.1). Ldp an d gida ed n h i l m vu md phdng, tim ra mdi lien he phi tuyln gida eic b i l n d i u v i o v i d i u ra eua van 64 daoc xem xet. Trong nghien edu niy, h i m ehdc nang sigmoid da dupe sddyng d l md hinh hda taong quan phi tuyln, dapc g i n v i o cic t l b i o t h i n kinh trong Idp I n , vdl phuong trinh nha sau:
I I I I
HJnft Z. 1: Ki^n Vdc cda md hinh mang no-ron nhin t^o (ANN) duoc sddung trong nghiin cOv niy Bin quan den
8 biin diu vio vi mot bidn diu ra Sd lapng t l b i o t h i n kinh tdi Uu trong ldp I n phu thudc v i o tang bai t o i n cu t h i . N l u sd laong t l bao t h i n kinh n h i n tao trong Idp an q u i Idn hoic qua nhd, nd
cd t h i se i n h hadng d i n hieu q u i eua vile da b l o [17], Trong nghiin edu niy, sau nhieu t h d nghilm, 8 t l bao than kinh cho Idp I n b i n g vdi kich thadc eua Idp dau v i o d i dapc chpn, phu hpp vdi sd dupe de xuat Wri Behnood v i Golafshani [18]. Cudi eung, thuat t o i n Levenberg Marquart d i dupe chpn cho q u i trinh dao tao md hinh ANN do hilu q u i cao (19].
2.3. ThuSt t o l n gili thuat di truyen (Genetic atgorithm)
Giii t h u i t di truyin (Genetic Algorithm - GA) I I ky thuit g i l i quylt b i i t o l n b i n g cich md phdng theo sd t i l n hda cua eon ngadi hay cua sinh v i t ndi chung (dya t r l n thuylt t i l n hda con ngadi cua Darwin) trong dieu kien mdi trddng sdng ludn thay ddi. Thuat g i i i di truyin l i mdt hadng t i l p c i n tinh t o l n g i n dung, nghia la myc t i l u eua t h u i t g i i i di truyin khdng n h i m dUa ra ldi g i i i chfnh xic tdi au m i I I daa ra Idi g i i i taong ddi tdi au. Ly thuylt n i y do Johm Henry Holland d l x u i t v i o gida t h i p niln 70 cua t h i ky XX. Thuit g i l i di truyen v l b i n chat l i t h u i t t o l n tim kiem dya theo quy l u i t cua q u i trinh t i l n hda t y nhien. Giii thuat k i t hpp sa sdng sdt eua e i u true khde n h i t trong sd cle cau true b i l u dien eie nhilm sic the vdi mdt sy trao ddi thdng tin dapc laa chpn n g l u nhiln d l tao t h i n h mdt t h u i t t o i n tim kiem.
Thuit t o i n g i i i t h u i t di truyin l i vi^c tinh t o i n tdi au hda, sd dung cic b i l u d i l n nhi p h i n v l eic so dd d l md hinh hda sd chpn Ipc, lai g h i p v i dot bien. GA d i dupe i p dyng thanh cdng cho nhieu linh vyc nhutdi Ou hda [20], hpc miy, mang t h i n kinh [21] v i bd d i l u khiln logic md [22]. Giii t h u i t di truyin sir dyng ba t o i n tCf sau diy:
• Lya chon: Mdt k l hoach lya chpn dapc i p dung d l xle djnh eieh eie e i t h i dUdc chpn de giao phdi daa tren khi ning tdn tai v i sinh sin trong mdt mdi tradng. Lya chon tao ra d i n sd mdi t d the h i cu, do do b i t d i u mdt t h i he mdi. Mdi nhiem sic the daoc d i n h g i l trong the h i h i l n tai de xic dmh g i l tri the lae eua nd. G i i trj the lue niy dapc sd dung de chpn cic nhiem sic t h i tdt han til quin the cho the he t i l p theo.
- Sa trao ddi eh4o hoic tai td hpp: Sau khi laa chpn, hoat ddng trao ddi chio dUdc i p dung cho eic nhiem sle t h i dapc chpn. Nd l i l n quan den viec h o l n ddi gen hole chudi trong ehuoi gida hai c i the. Q u i trinh niy dapc lap 1^1 vdi cic e l the eha me khie nhau cho d i n khi the he t i l p theo cd du cAc c l the. Sau khi trao ddi cheo, t o l n t d dpi bien dupe i p dung cho mdt tap hop con dupe chpn ngiu nhiln eiia q u i n t h i .
- Ddt biln:€)dt b i l n l i m thay doi nhiem sic the theo nhdng eieh nhd de daa ra nhdng d i e diem tdt mdi. Nd duoc I p dyng de mang lai sada dang trong d i n sd.
Trong nghien edu niy, t h u i t t o i n giai t h u i t di truyin dapc SLf dung de tdi uu hda v i tim ra trpng sd, dp lech cho cac te b i o t h i n kinh trong Idp an cua mang ANN.
2.4. D i n h g i l k h i ning dU b l o cua mo hinh De danh g i i k i t q u i d u b i o do cic md hinh AI dda ra, hai t i l u chi ddpc sif dung trong nghiin edu nay la he sd taong quan (R) {correlation coefficient) v i sai sd RMSE (Root Mean Square Error).
KHOA HOC CONG NGHE
SAO(S2020
R la mdt chi t i l u quan trpng trong p h i n tlch hdi quy.
Chi sd R b i l u dien taong quan gida k i t q u i da d o l n v l d i u ra thae t l , cd g i i tri thay ddi t d 0 d i n 1. G i l tn cua R cang cao thi cang bieu thi mdi taong quan tdt gida g i l tri d a d o i n va g i i tri thac t l . Chi sd RMSE I I mdt phucmg p h i p xic dinh sai sd khic, daa tren chenh lech binh phaong trung binh gida d i u ra dUOe dU d o l n v l d i u ra thac t l . Gia tri RMSE t h i p cho thay hieu suit tdt hdn cua t h u i t t o i n A!. Cdng thdc tfnh R v l RMSE cd t h i daoc tham khio trong cic t i i lieu trfch dan [23].
3. K ^ QUA VA TH AO L U A N
Co sd dd lieu chda 425 sd lieu dapc chia t h i n h hai t i p coniTip h u l n luyin (70% d d l i l u ) v l tap k i l m chdng (30% d d lieu edn lai), duoc chpn ngau nhien t d eo sd dQ lieu ban d i u theo h i m xic suit p h i n phdi ehuln. Mic du eic ty le d i o tao/kilm chdng khic nhau cd t h i dUde SLf dung. Trong b i i b i o niy, ty II70/30 cho eie bd d d l i l u h u l n luyen/kiem chdng da dupe chpn, nha daoc d l xuit trong nghien edu [24]. Phin tieh thdng k l d i dape thuc h i l n d l chdng minh ring su lUa chpn eiia cic tham sd d i u v i o khdng ed mdi taong quan cheo nao gida eddng dp nen cua HPC v l 8 b i l n dau v i o [13-16]. Hai md hinh ANN dape d l x u i t l i md hinh ANN sd dung thu§t t o i n Levenberg Marquart v i md hinh ANN dapc tdi Ou hda b i n g t h u i t t o i n g i i i t h u i t di truyin. Cau true md hinh gdm ba Idp vdl 8 no-ron trong mdt ldp I n duy nhit da dupe I p dung trong c l hai mdhinh.Tatcl cle gia tri eua b i i t o i n , bao gdm d i u v i o v i h i m mue tieu, dape chuan hda ve kholng g i i trj tir 0 d i n 1 de g l i m thieu sai sd do md phdng tao ra. Trong cic b i i t o i n da b l o ndi chung, n i n g lac da b l o cua md hinh l i quan trpng nhit. Nd dupe the hien thdng qua cie chi t i l u d i n h g i i sai sd, nha da trinh biy d phan trade.
Hinh 3.1 vd 3.2 cho thIy g i i tri eddng do n^n eua cie thi nghiem thu ddpc (dddng lien tyc) v i cle g i l tri da d o i n (khdng lien tuc) t d tap h u l n luydn v l tap k i l m chdng eiia hai md hinh ANN va ANN-GA. Cadng dp n I n da d o i n cua 298 mau trong t i p h u l n luy^n cua c l hai md hinh d l xuat rat khdp vdi cic k i t q u i eiia md phdng.
Vdi tap d d lieu kiem chung, 127 k i t q u i thi nghilm cung dupe da b i o vdi sai sd nhd qua c i hai cdng cy ANN v i ANN-GA. Cu t h i , cic sai sd v i tdong quan gida k i t q u i thf nghiem v i md phdng se dapc d i n h g i i d phan t i l p theo.
Mau thi nghiem
39
S6 06.2020
j:Pi;;,fi#
D 50 100 150 200 250 300 Mau thi n g h i # m
Hinh 3.1: Cie gii tri tN nghiem vi m6 phdng v4 eudng d6 nin cua HPC 28 ngiy vdi m6 hinh ANN vi ANN-GA cho tip huA> luyin
' i \ U
Jj i f nil;
_ _ Ho pfiong ANN 20 40 60 SO 10O 120
Mau t h i n g h i f f n
Thi nghiem Mo phong ANN-GA 40 EO
M a u ttii n g h i ^ Hinh 3.2: Cie gii tri thi nghUm vi m6 ptidng ve eudng do nen cua HPC 28 ngiy vdi mo hinh ANN vi ANN-GA cho tip kiem chung
Hinh 3.3. vd 3.4 t h i h i l n mat do xle suit sai sd cho t i p h u l n luyin v l t i p kiem chdng eua hai md hinh ANN v l ANN-GA. Phin tieh sai sd cho thIy, ddi vdi p h i n h u i n luyen, e l hai md hinh AI d i u mang lai k i t q u i tdt trong pham VI hop ly (nghTa I I , cd r l t it mau cd sai sd Idn han 10 MPa v i sai sd chd y l u trong pham vi ±4 MPa). Tuy nhien, md hinh ANN-GA ed pham vi sai sd nhd hon, cho thIy md hinh niy ed n i n g lac da t>lo tdt hon. Tuong ta, ddi vdl phan kiem chdng, c i hai md hinh ANN va ANN-GA d i u cung c i p ket qua d d d o l n tdt, dua tren cac phin tich thdng ke cua sai sd (Hinh 3.4}. Die bilt, md hinh ANN-GA cho sai sd thap hon h i n md hinh ANN truyen thdng.
SAI \6 trung b«^h • 1 - ^ Do trch chuln > 7 1114 25 Sal %b RMSE •
Sai SO (MPa) H W i 3.3- Mat dd xic suat sai so cho tap huan luyen
vdl md hinh ANN vi ANN-GA S»l s6 irung binh = 1 8296 ' Ool¥Chchu4n = 7 1164 J
RMSe = T 3206 |
S a i s o ( M P a ) Sal th (runs bmh = 0 512S4
€>pl*chchuin-&413l : Sal t& RMSE • S.41
li | j .
jj
S a i s d ( M P a ) Hinh 3.4: Mit dd xic suSt sai sdctto tip kiem chung
vdi md hinh ANN vi ANN-GA
40
KHOA HOC CONG NGHE
56 06/2020
Hinh 3.5 vd 3.6 cho thay, hai md hinh AI d l xuat ed mdi taong quan tuyen tinh manh gida c i c g i l ^ eadng dd n I n dupe d a d o i n va cadng dd nen thae te. Hieu q u i da b l o cic g i l trj cadng dd n I n eiia dd l i l u h u l n luyen v l k i l m chdng l i tdt, vdi R = 0,86765 v i R = 0,8959 cho thuit t o l n ANN; v l R = 0,89233, R = 0,91525 cho t h u i t t o i n ANN-GA. Mdt Ian nda, c i c g i i tri eadng dp nen cua HPC 28 ngiy tudi ed the daoc ANN v i ANN-GA d a d o l n thinh cdng vdi cic sai sd n i m trong kholng g i l tri hop ly. Nhdng quan sit nhd v i y da chi ra ring, hai thuat t o l n niy I I hdu h i l u v i r l t cd t i l m nang de dd d o l n g i i tri cadng 66 n I n eiia HPC 28 ngiy. Vi ANN-GA da i p dung thuit t o i n g i l l t h u i t di truyin de tdi au hda cau true eua ANN, n I n trong trddng hop niy, h i l u q u i d g b i o cua ANN-GA tdt hon ANN truyen thdng.
80
/U
eo m
in
30 20
0 Dt> l i ^ u R = 0 8 6 7 6 5 Y = T
» oo
» o * ' » « . 0 ^ ^
^b - ' V ^ ^ ^ J A / ' '
...- . • . f$Wi >-''
'^*/'
"
*0 / "
'•A^J^O
»r ^ "
» "
20 40 60 f t h i n g h i e m ( M P a )
0 Ou> l i ^ u R = 0 . 8 9 2 3 3
Y = T 5 0
y
„ " oS> ^
" "^^y
0 X * y<^x
!$> S\-r^.^/^' •^
^ • • ^ l ^ ' ' ...m^s,:
• . y « ? - ° "^ 5
'ifyfiS^''^ -
J ^
•^ •
i> /
?
• \
r t h i n g h i e m ( M P a ) Hinh 3.5: Ket qui tuong quan eua cie gii tri eudng dp nen
thuc t^vi dudoin cho phSn huan luyen cua md hinh ANN vi ANN-GA
7 0
6 0
5 0
4 0 Di> lieu R=0.895g Y = T
}
°
2 0 4 0 f t h i nghiem (MPa)
70 60 SO' 40 30
9 Dirli^u
R = 0 , 9 1 5 2 5 : •>
» ' » J *,•
' ^/^
3 Z ' , jf ' *
f ^ t h i n g h i # m (MPa) Hinh 3.6: Ket qui tuong quan ciia cic gii tri cudng dp n
thuc tevi dudoin cho phan kiem chdng cua md hinh ANN vi ANN-GA
4 . K^T L U A N V A K I I N N G H !
Mang ladi than kinh n h i n tao cd k h i n i n g hpc hdi va khli q u i t hda tCr cic dd lieu, dieu n i y lam cho mang ladi than kinh nhan tao trd thanh mdt cdng ey manh me de g i i i quyet cac sd van d l ky t h u i t phdc tap. Trong nghiin edu nay, tan dyng eic du diem cua mang ladi t h i n kinh n h i n tao d l d u d o i n eadng dd n I n cua HPC 28 n g i y m i khdng c i n dung bat ky thi nghiem n i o vdi hai k i l n true mang no-ron n h i n tao I I ANN v i ANN-GA.
Hai md hinh AI ddpc p h i t trien vdi cau true mang (8-8- 1]. Khi n i n g d a d o i n eua eie md hinh Ai daoc d i n h g i i b i n g hai tieu chf l i he sd tdong quan R va sai sd RMSE.
K i t q u i cho thay cic md hinh Ai daoc phat tnen trong nghiin edu n i y d i thac hien tdt viec dd d o l n cadng do n I n HPC 28 ngiy. nhdng ANN-GA {R = 0.91525, RMSE = 6.41) cho h i l u suit tdt hon so vdi ANN. Do do, ed the k i t l u i n ring, eie md hinh AI l i mdt phuong p h i p day hda
41
SA Ob 20 20
hen de du d o i n caong dp nen cua HPC ngoat ra co t h i duoc i p dung d l du doan cie ttnh chit quan trpng khae cua HPC nha cadng dd klo. eadr>g dd udn hoic md-dun dan hdi. Nhin chung, nghien edu niy giup cho viee lya chon cic md hinh AI phu h<^ de xac dinh nhanh cac tfnh c h i t CO hoc cua HPC.
Tli lieu tham khio
11]. E. Nawy (2001), Fundamentals of High- Performance Concrete, volJohn Wiley, no NJ. USA.
[2]. R-K Chang and Y.-N. Peng (2001), Influence of mixing techniques on properties of high performance concrete. Cem. Coner. Res., vol31, n a l , pp.87-95, doi:
httpsJ/doi.org/10. }0i6/SOOO8-8846(00)00439-7.
[3].V.Q.Tranetal. (2019), Effect o^temperafureonf/ie chloride binding capacity of cementitious materials, Mag.
Coner Res.,pp.l-t4,Decdoi: 10 1680/jmacr.l9.00484 [4J. F. P. Zhou, F. D. Lydon, and B. I. G. Barr (1995), Effect of coarse aggregate on elastic modulus and compressive strength of high performance concrete. Cem.
Coner. Res., vol.25, no.1, pp.177-186, doi: httpsj/doi.
org/10.1016/0008-8846(94)00125-L
[5]. R, Duval and E. H. Kadri (1998), Influence of Silica Fume on the Workability and the Compressive Strength of High-Performance Concretes, Cem. Coner. Res., vol^8, no.4, pp.533-547. doi: https://doi.org/10.1016/S0003- 8846(98)00010^.
[6]. Y. N. Chan, X. Luo and W. Sun (2000), Compressive strength and pore structure of high-performance concrete after exposure to high temperature up to 80ffC, Cem.
Coner. Res., vol.30, n o ^ , pp.247-251, doi: httpsJ/doi.
org/10.1016/50008-8846(99)00240-9.
[7]. P. Rashid, M A ; Mansur, M A : Paramasivam (2002), Correlations between Mechanical Properties of High-Strength Concrete. J. Mater. Civ. Eng, pp.230-238.
18]. S. Shams, M. Monjezi, V J. Majd and D. J.
Armaghani (2015), Application of fuzzy inference system for prediction of rock fragmentation induced try blasting, Arab. J. Geosci., vol 8, no.l2, pp.10819-10832, doi:
10.1007/s12517-015-1952-y.
[9]. H. Q. Yang, S. G. Xing. Q Wang and Z. U (2018), Model test on the entrainment phenomenon and energy conversion mechanism of Row-like landslides. Eng. Geol., vol.239, pp.] 19-125, doi: https://doi.Org/iO.1016/j.
enggeoJOi 8.03.023.
[10]. A. Kaveh and M. Maniat (Jun 2014). Damage detection in skeletal structures dosed on charged system search optimization using incomplete modal data TT, UCE, voLl 2, na2. pp.193-200.
[11). H. Q. Yang, Y. Y. Zeng, Y. F. Lan and X. P. Zhou (2014), Analysis of the excavation damaged zone around a tunnel accounting for geostress and unloading. Int J.
Rock Mech. Mm. Sci., vol. -69, pp.59-66, doi: https-J/doi.
org/10.i016/j.ijrmms.2014.03.003.
[12], J. Zhou et al. (2019), Slope stability prediction for circular mode failure using gradient txtosting machine approach based on an updated database of case histories.
SatSci, vol.118,pp.505-518,doi:/inpiy/do/-Ofg 10.1016/)- ssd2019.05.046.
[13]. l.-C.Yeh (1998), Modeling concrete strength with augment-neuron networks, J. Mater Civ. Eng, pp J63-268.
[14]. I.<. Yeh (1999), Design of high-performance cofKrete mixture using neural networks and nonlinear programming. J. Comput. Civ. Eng, pp-36-42.
[15]. 1.-C. Yeh (1998), Modeling of strength of high- performance concrete using artificial neural networks, Cem. Coner. Res, pp. 1797-1808.
[161. I.-C. Yeh (2006), A mix proportioning methodology for fly ash and slag concrete using artificial neural networks, J. Mater. Civ. Eng, pp.597-604.
[171. A. K. Al-Shamin. J. H. Kim. T.-F Yuan and Y.
S. Yoon (2019). Modeling the compressive strength of high-strength concrete: An extreme learning approach.
Constr. Build. Mater, voL208. pp.204-219, doi: httprJ/doi.
org/10.1016/j.conbuildmat.2019.02.165.
[18]. A. Behnood and E. M. Golafshani (2018), PredKting the compressive strength of silica fume concrete using hybrid artificial neural network with multi-objective grey wolves. J. Clean. Prod., vol.202, pp.54-64, doi: https://
doi.org/10.1016/).jclepro.2018.08.065.
[19], D,-K. Bui.T. Nguyen, J.-S. Chou, H Nguyen-Xuan and T. D. Ngo (2018), A modified firefly algorithm-artificial neural network expert system for predicting compressive and tensile strength of high-performance concrete. Constr.
Build. Mater., vol.180, pp,320-333, doi: https-J/doi.
org/10.l016/j.conbuildmat.2018.05.201.
[20], D.T Pham and D Karaboga (Mar. 1997), Gene(/c algorithms with variable mutation rates: Application to fuzzy logic controller design, Proc. Inst. Mech. Eng. Part I J. SysL Control Eng., vol.2n, no.2, pp.157-167, doi:
10.1243/0959651971539975.
[21]. L B. Booker, D. E. Goldberg and J. H. Holland, Classifier systems and genetic algorithms. Artif. Intetl., vol.40, no.l, pp.235-282, 1989, doi: httpsJ/doi.
org/10.10l6/0004-3702(89)90050-7.
[22]. M. A. Lee (1993). Dynamic Control of Genetic Algorithms using Fuzzy Logic Techniques Visiting Industrial Fellow from Central Research. Evaluation.
[23]. H. B. Ly et al. (2019), Prediction and sensitivity analysis of bubble dissolution time in 3D selective laser sintering using ensemble decision trees. Materials (Basel), vol.12, no.9, doi: 10.3390/ma12091544.
[24]. M. S. Khorsheed and A. O. Al-Thubaity (2013), Comparative evaluation of text classification techniques using a large diverse Arabic dataset. Lang Resour Eval., vol 47, no.2.pp.513-538,doi: 10.1007/510579-013-9221-8.
Ngiy nhin b i i : 30/3/2020 Ngay chip nhin d ^ n g : 02/5/2020 Ng J«i phin bi#n: TS. Mai Thj H i i V l n
TS. T l i n Van Q u i n