Professional Documents
Culture Documents
4
2019
4 ACTA AUTOMATICA SINICA April, 2019
êâ°Ä2ÀL§$1")Í{
ñ² 1
[å 1
_ 1
ÆUn 1
Á 2ÀL§´|^¶ÔY½Õí5½²J?nY½Õí5?1¶Ô©lÔnL§. ©Ï
Lïᱶó Ú¶ó6þǑÑ\, ±2ÀL§°¶¬ ¶¬ ǑÑÑõCþ!rÍÜ!5!C$1L§
., |^ïÄ
ûÿA:, JÑ
)¶Ô¬ $1L§ì°Ä.!PID ì!")Íì!
ïÄÖìêâ°ÄÚ`ïÄÖ PID )Í{, ¢y
ØØ!·)ÍïÄ
Ö, ÏL2ÀL§$1"ý¢y
©¤J{k5.
'
êâ°Ä, 2ÀL§, $1, )Í
Ú^ª ñ², [å, _, ÆUn. êâ°Ä2ÀL§$1")Í{. gÄzÆ, 2019, 45(4): 759−770
DOI 10.16383/j.aas.2018.c170552
Abstract The flotation process is a mineral separating physical process by taking advantage of the hydrophilic or
hydrophobic properties of the mineral or the hydrophilic or hydrophobic properties obtained by treatment. In this
paper, firstly, a multivariable, strong coupling, nonlinear and time-varying operational process model is established with
the input and output of the pulp level and feed flow as its inputs and the concentrate grade and tailing grade as its
outputs. Secondly, by taking the advantage that the unmodeled dynamics at last sampling point can be measured,
a scheme of one-step optimal unmodeled dynamic compensation PID decoupling control is proposed including the ore
grade operational process controller driven model, PID controller, feedback decoupling controller and unmodeled dynamic
compensator, to guarantee zero steady-state error, static decoupling, and unmodeled dynamics compensation. Finally, a
simulation experiment on the operational feedback control in an industrial flotation process is conducted to demonstrate
the effectiveness of the proposed method.
Key words Data-driven, flotation processes, operational control, decoupling
Citation Jiang Yi, Fan Jia-Lu, Jia Yao, Chai Tian-You. Data-driven flotation process operational feedback decoupling
control. Acta Automatica Sinica, 2019, 45(4): 759−770
2ÀL§£ã 2ÀL§Ä.©Û
1.1.1
1.1
¶6þÚYópÝ´±¶zmÝ u (t)
2ÀL§´|^¶ÔY½Õí5½ ¶¶ó"= u (t) ǑÑ\, YópÝ y (t)
1
ÔnL§. ©±Ô¶ü2Àø2ÀL§Ǒ~,
2
TL§ÌdøN!¶z!°¶z!¶z y2 (t) k1 p
ẏ1 (t) = − u1 (t) 2gy1 (t) (1)
!Ïí86ì|¤, Xã 1. A A
x(t) = F (x
ẋ x(t), y (t)) ÅÄ Ñ¬E¤2Àø¥ÔÆpÇX IQ (t) ,
ê )Cz Ó, duTÔnCz kp1 , kp2 , ke1 , ke2 ,
E, ¶â2Àø?í86ì>6 I (t) éÔ
x (t))
r (t) = G(x (3)
,
Ù¥, G x(t) = [x (t), x (t), x (t), x (t)] , ǑY ÆpÇXê )KǑJ±^ê
Q
T
kp1 , kp2 , ke1 , ke2
ó¥¶Ôþ!óþ& ¥¶Ôþ! Æ.£ã
du2À
óSßL§É
1 2 3 4
, ,
óþ; Ñ\ y (t) = [y (t), y (t)] , Ǒ2Àø À¶¶ó£66Ä Ó¶¶Ô¬ ÅÄ, T
,
¶6þYópÝ; ÑÑ r (t) = [r (t), r (t)] 7,Ô¶óßÝ ó¶óßÝ X
1 2
T
Xa1 2
©OǑ2ÀL§ó:?éAG!
x eq , y eq , r eq
1
x3 gcp 2
+ x4 gcp
Ñ\!ÑÑ. o (·) Ú o (·) Ǒ55z
L
p¡þãÅCz)ïÄ.
1 2
x3 + x4 cu
x(t)) =
G(x (5)
éAæǑǑ T (æ±ÏǑ T = mT ,
x1 g1 + x2 g2
T Ǒ æ±Ï, m Ǒê) lÑ.Ǒ
cp cp o d
Lcu
x1 + x2 d
ª (4) ¥, X = (g − g )/(g − g )X .
2 1 2 1
x(T ), y (T ))
x (T + 1) = Adx (T ) + Bdy (T ) + o 3 (x
. (4) . (5) ¥ëê¹ÂXL 1 ¤«.
a cp a a cp a
x(T ) + o 2 (x
r (T ) = Cx x (T )) (7)
L 1 2ÀL§ÎÒL Ù¥ ǑlÑ^ RT
, Ad = eĀTo , Bd = 0 o eĀt dtB. o 3 (·)
Table 1 Flotation process symbol table eïÄ .
ÎÒ Ôn¹Â ÎÒ Ôn¹Â K¶ÔL§Ñ\ÑÑé.Ǒ
kp1 Ô¶2ÀÇ kp2 ó2ÀÇ A(z −1 )rr (T + 1) = B(z −1 )yy (T ) + v (T ) (8)
ke1 Ô¶üÇ ke2 óüÇ Ù¥,
ga ¶óÔ¶¬ Xa1 Ô¶óßÝ
A(z −1 ) = diag{A1 (z −1 ), A2 (z −1 )}
Xa2 ó¶óßÝ A 2Àø.¡È " #
2ÀøpÝ ¶6þ B (z ) B (z )
−1 −1
H qT 11 12
B(z −1 ) =
Lcu Ô¶¶Ô¬ qc °¶6þ B21 (z −1 ) B22 (z −1 )
Ô¶ó ó¶ó
" #
v11 (T )
1 2
v (T ) =
Ô¶¬ Ô¶¬
gcp gcp
v12 (T )
B11 (z −1 ) = b11 11 −1
L§¬)¶â2Àø?í86ì>6
1 2
B21 (z −1 ) = b21 21 −1
0 + b1 z
762 g Ä z Æ 45 ò
B22 (z −1 ) = b22 22 −1
0 + b1 z 2 $1")ÍìO
2ÀL§$1üÑ
L«ïÄ.
2.1
v (T )
éu¢S2ÀL§, du$1÷v d¶6þÄåÆ., ¶6þL§Ǒ
½Ônå, ¶6þ÷v¶"óÇ 5L§, ±ÏL PI ì¢y½,
Ônå, ¶ópÝ÷v2ÀøpÝå. Ó dYópÝÄåÆ., YópÝL§´±
, °¶¬ ¶¬ ÷v½, = 0 % ∼ ¶6þǑÿ6Ä5L§, ±ÏL
"
100 % m, 3(½
XÚÑ\ÑÑëê A(z ) Ú
PI ì¢y, ©b2ÀL§.
´½, ¶6þYópݱ½/l
−1
«. Tì)¶Ô¬ ì°Ä.!PID
Ù¥, L Ú L ©O´°¶¬ 8I ì!")Íì!ïÄÖì. ½
þe, L Ú L ©O´¶¬ 8I Ǒ8I«m¥, r = (L + L )/2, r
cg min cg max
* *
þe.
tg min tg max 1 cg max cg min 2
= (Ltg max + Ltg min )/2.
(14), Ú\Xe5UI.
∗ −1 −1
Ä.ÑÑ, L«Ǒ
J =
P (z −1 )rr (T + 1) − R(z −1 )rr ∗ (T ) +
r̄r (T ) = −A (z ∗ −1
)rr (T ) + B(z −1
)yy (T − 1) (12)
Q(z −1 )yy (T ) + S(z −1 )yy (T ) +
du v (T − 1) ±dª (12) ¼, æ^Xã 2
¤«êâ°ÄÚ`ïÄÖ PID )
2
(16) K(z −1 )vv (T − 1)
Í(é v (T − 1) ?1ÄÖ PID )Í Ù¥, P (z ), Q(z ), R(z ), K(z ) Ǒ'u z
Æ, =
−1 −1 −1 −1 −1
éÆõªÝ
, S(z ) Ǒ'u z éÆǑ" −1 −1
õªÝ
.
y (T ) = y (T − 1) + K [ee(T ) − e (T − 1)] +
K e (T ) + K [ee(T ) −
I D
Ú\2ÂÑÑ φ(T + 1) Ǒ
P
Ø, K , K , K ǑéÆ~êÝ
, ©OL« e (T + 1) = φ(T + 1) − φ (T + 1) = ∗ ∗
PID ì'~!È©Ú©XêÝ
. G(z ),
P I D
−1
P (z )rr (T + 1) − R(z )rr (T ) + −1 −1 ∗
õªÝ
, G(z ), H(z ), K̄(z ) Ǒ'u z K(z )vv (T − 1) −1
−1 −1 −1 −1 (19)
éÆõªÝ
, Ù¥ G(z ) = (1 − z )K + Ú\Xe Diophantine §
−1 −1
P
K + (1 − 2z + z )K , H(z ) = diag{1 − z ,
−1 −2 −1 −1
Ý
.
- B(z ) = B̄(z ) +¯ B̄¯ (z ), B̄(z ) Ǒ' Ù¥
−1 −1 , F (z ) Ḡ(z ) gǑ n = 0, n = 2,
−1 −1
−1 −1
F Ḡ
u z éÆõªÝ
, B̄(z ) Ǒ'u z é Ḡ(z ) = Ḡ + Ḡ z + Ḡ z . −1 −1 −2
ÆǑ"õªÝ
. òÆ (14) \é kH(zd Diophantine §, J (J =
−1 −1 −1 0 1 2
kïÄÖÚ
min
(8), 4é§.
2
v
)∆v (T )k ) −1
`ÆǑ
H(z −1 )A(z −1 ) + z −1 B̄(z −1 )G(z −1 ) r (T + 1) =
B̄(z )G(z )rr (T ) + H(z )∆vv (T ) +
−1 −1 ∗ −1 F (z −1 )B̄(z −1 ) + Q(z −1 ) y (T ) =
h i
¯ (z −1 ) − B̄(z −1 )H̄(z −1 ) y (T ) + R(z −1 )rr ∗ (T ) − Ḡ(z −1 )rr (T ) −
H(z −1 )B̄ h i
S(z −1 ) + F (z −1 )B̄¯ (z −1 ) y (T ) −
H(z −1 ) − B̄(z −1 )K̄(z −1 ) v (T − 1) (15)
Ù¥, ∆vv(T ) = v (T ) − v (T − 1) ǑïÄC
F (z −1 ) + K(z −1 ) v (T − 1) (21)
K = ΠḠ
«, æ^Æ (14), 4XÚÑ\ÑÑ&Ò
D 2
K̄(z ) = Π F (z ) + K(z )
−1
(26) ½ ),
−1
= −1
d½Â (24) , Æ (14) ÷vXe: kyy (T )k < ∞, krr (T )k < ∞ (31)
¯ (1) = B̄(1)S(1)
Q(1)B̄
Øuýk½ ε, =
Q(1) = B̄(1)K(1) (27) lim krr ∗ (T ) − r (T )k ≤ ε (32)
T →∞
=¢y
ØØ!·)ÍïÄ y². ÏǑïÄ v (T ) k., ¤±
Ö. ïÄCz ∆vv (T ) = v (T ) − v (T − 1) k
4 Ï ñ²: êâ°Ä2ÀL§$1")Í{ 765
, r (T ) ∆vv(T ) k.. Kk
∗ M = 16.8 t/m , p
1 3
Mp2 = 1 123 t/m3
Me1 = 4.56 t/m3 , Me2 = 0.2 t/m3
k∆vv (T )k = kvv (T ) − v (T − 1)k ≤
y1 (t) = 2.6 m, y2 (t) = 17 m3 / min
kvv (T )k + kvv (T − 1)k ≤ 2M (33)
d v (T ) r (T ) k.5, K3~ê d , æ±ÏǑ T
∗ o = 30 min, Xe.:
· · · , d . æ^©z [20−22] aq{, ±y²
1
4 A(z ) = −1
"
kyy (T )k ≤ d1 + d2 M < ∞ (34) 1 − 0.1767z −1 + 0.0038z −2
0
krr (T )k ≤ d3 + d4 M < ∞ (35) #
=÷v4XÚÑ\ÑÑ&Òk. (BIBO 0
'ý¢. F (z −1 ) = I2×2
4.2 Äu5.ýé'¢
Ḡ(z ) =
−1
,
©{IÑÑêâ¢y".
" #
4.6014 × 10 0 −4
(41)
0 2.3477
KI = Π(Ḡ0 + Ḡ1 + Ḡ2 ) =
" #
0.0212 0
(42)
0 108.1049
" #
0 0
KD = ΠḠ2 = (43)
0 0
2) .ýÿ{¥5UI
T +Np
T
X
J= (rr ∗ − r (i)) Q(rr ∗ − r (i))
i=T +1
{¥ýÿÚǑ N = 5 Ú, \Ý
Q
Ǒü Ý
, `z¯K)^ MATLAB ¥
p
ã 4 5.e PID )ÍÑ\
quadprog ¼ê. Fig. 4 Control input curve with PID decoupling control
ý¢¥½8I«mǑ under linear model
4 Ï ñ²: êâ°Ä2ÀL§$1")Í{ 767
4.3 Äu5.ýé'¢
Ǒ
y©{¢SJ, ^©z [15]
¥5é?1ý¢, =2ÀL§.
(3), ^©êâ°ÄÚ`ïÄÖ
PID )Í{.ýÿ (Model pre-
dictive control, MPC) {?1é'ý¢, Ó
, Ä3¶¬ ëê g Å6Ä, ëê
6ÄXã 8 ∼ 12 ¤«.
a
Ǒ
µd©J, Ú\ýéØÈ
© (Integral absolute error, IAE) Øþ
(Mean square error, MSE) 5µdJ,
[23−26]
ã 5 5.e.ýÿ¶Ô¬ l
Fig. 5 Ore grade tracking curve with MPC under
linear model
ã 8 5.eêâ°ÄÚ`ïÄÖ
PID )ͶԬ l
Fig. 8 Ore grade tracking curve with data driven one
step optimal unmolded dynamic compensation PID
decoupling control under nonlinear model
ã 6 5.e.ýÿÑ\
Fig. 6 Control input curve with MPC under linear model
ã 9 5.eêâ°ÄÚ`ïÄÖ
PID )ÍÑ\
Fig. 9 Control input curve with data driven one step
ã 7 ëê6Ä optimal unmolded dynamic compensation PID decoupling
Fig. 7 Parameter disturbance curve control under nonlinear model
768 g Ä z Æ 45 ò
úªǑ
∗
T
X
IAE = |ri∗ − ri (T )|, i = 1, 2 (50)
T =1
v
u
u 1 X T∗
2
M SE = t |r ∗ − ri (T )| , i = 1, 2 (51)
T ∗ T =1 i
µdIXL 2 ¤«.
L 2 é'¢µdI
Table 2 Performance index of comparison experiment
lý(J±wÑ, éǑ5
., ©¤J{±UC¶6þ
pÝ, ¶Ô¬ l8I½. ©z [15] ¥
.ýÿ{{â"êâNÑ\,
¶Ô¬ l8I½. Ó§ÑÑ8I
3½{, ©JÑ{µd5UI
`u©z [15] ¥.ýÿ{.
5 (Ø
ã 11 5.e.ýÿ¶Ô¬ l ©JÑ2ÀL§êâ°ÄÚ`ï
Fig. 11 Ore grade tracking curve with MPC under ÄÖ PID )Í{, d¶Ô¬ ì
nonlinear model °Ä.!PID ì!")Íì!ïÄ
Öì|¤, ¢y
ØØ!·)Í
ïÄÖ, ü2ÀøǑéý¢(J
L²
T{k5.
References
1 Chai Tian-You. Operational optimization and feedback con-
trol for complex industrial processes. Acta Automatica
Sinica, 2013, 39(11): 1744−1757
(ÆUn E,óL§$1`z" gÄzÆ
. . , 2013,
39(11): 1744−1757)
4 Yang Ya-Ru, Li Shao-Yuan. Economic model predictive con- 16 Jiang Y, Fan J L, Chai T Y, Chen T W. Setpoint dynamic
trol for global optimal operation of nonlinear switching sys- compensation via output feedback control with network in-
tems. Acta Automatica Sinica, 2017, 43(6): 1017−1027 duced time delays. In: Proceedings of the 2015 American
æT o 5XÚÛ`z$1²Lýÿ
( , . . Control Conference (ACC). Chicago, IL, USA: IEEE, 2015:
gÄzÆ , 2017, 43(6): 1017−1027) 5384−5389
5 Zhang Xiang-Yu, Li Ji-Geng, Zhou Ping, Zhang Zhan-Bo, 17 Wang T, Gao H J, Qiu J B. A combined fault-tolerant and
Liu Huan-Bin, Wang Hong. Operational optimization con- predictive control for network-based industrial processes.
trol system for pulp chlorination process. Control Engineer- IEEE Transactions on Industrial Electronics, 2016, 63(4):
ing of China, 2014, 21(2): 303−308
Ü oU ±² ÜÓÅ 4Q ÷ óŤL§$1
2529−2536
( , , , , , .
`zXÚ ó§ . , 2014, 21(2): 303−308)
18 Fan J L, Jiang Y, Chai T Y. MPC-based setpoint com-
pensation with unreliable wireless communications and
6 Liu Xiao-Qing, Cheng Quan, Li Jin, Zhou Xiao-Dong. In-
constrained operational conditions. Neurocomputing, 2017,
tegrated automation system for flotation processes. Control
270: 110−121
Engineering of China, 2016, 23(11): 1702−1706
4¡ § oA ±À 2À)L§nÜgÄzXÚ
( , , , . .
ó§ , 2016, 23(11): 1702−1706) 19 Jury E I. Inners and Stability of Dynamic Systems. Malabar,
Florida, India: Krieger Pub Co, 1982.
7 Wang R H, Qiu M J, Zhao K L, Qian Y. Optimal RTO
timer for best transmission efficiency of DTN protocol in 20 Chai Tian-You. Multivariable Adaptive Decoupling Control
deep-space vehicle communications. IEEE Transactions on and Its Application. Beijing: Science Press, 2001.
Vehicular Technology, 2017, 66(3): 2536−2550 (ÆUn õCþg·A)Í9A^ ® ÆÑ
. . : , 2001.)
8 Pan Hong-Guang, Gao Hai-Nan, Sun Yao, Zhang Ying, 21 Chai T Y, Zhai L F, Yue H. Multiple models and neural net-
Ding Bao-Cang. The algorithm and software implementa- works based decoupling control of ball mill coal-pulverizing
tion for double-layered model predictive control based on systems. Journal of Process Control, 2011, 21(3): 351−366
multi-priority rank steady-state optimization. Acta Auto-
matica Sinica, 2014, 40(3): 405−414
( ù1 p°H Ü= ¶ñ Äuõ`k?`zV
, , , , . 22 Jia Yao, Yue Heng, Chai Tian-You. Multi-operation condi-
(ýÿ{9^¢y gÄzÆ . , 2014, 40(3): 405−
tion switching control for high pressure acid leaching pro-
cess. Control Theory and Applications, 2014, 31(10): 1318−
414)
1326
_ ð ÆUn pØEL§õó¹{ nØ
( , , . .
9 Ding J L, Modares H, Chai T, Lewis F L. Data-based multi-
objective plant-wide performance optimization of industrial
A^ , 2014, 31(10): 1318−1326)
processes under dynamic environments. IEEE Transactions
on Industrial Informatics, 2016, 12(2): 454−465 23 Hägglund T. A control-loop performance monitor. Control
Engineering Practice, 1995, 3(11): 1543−1551
10 Wang T, Gao H J, Qiu J B. A combined adaptive neural
network and nonlinear model predictive control for mul- 24 Jia Yao, Chai Tian-You. Interval cascade intelligent control
tirate networked industrial process control. IEEE Trans- in vaper-water plate-type heat exchange process. Acta Au-
actions on Neural Networks and Learning Systems, 2016, tomatica Sinica, 2016, 42(1): 37−46
27(2): 416−425 _ ÆUn ðYª9L§«mG?U{ gÄz
( , . .
Æ , 2016, 42(1): 37−46)
11 Yang X P, Chen Y R. Intelligent control and optimization of
the coal slime flotation. Advanced Materials Research, 2012, 25 Jia Y, Chai T Y. A data-driven dual-rate control method for
524−527: 1007−1010 a heat exchanging process. IEEE Transactions on Industrial
Electronics, 2017, 64(5): 4158−4168
12 Li H B, Chai T Y, Zhang L Y. Hybrid intelligent opti-
mal control for flotation processes. In: Proceedings of the
26 Wang Lan-Hao, Jia Yao, Chai Tian-You. Dual-rate interval
2012 American Control Conference (ACC). Montreal, QC,
control of pump pool level and feeding pressure during re-
Canada: IEEE, 2012. 4891−4896
grinding. Acta Automatica Sinica, 2017, 43(6): 993−1006
=Í _ ÆUn 2L§"³ Ú¶ØåVÇ«
( , , .
13 Jiang Y, Fan J L, Chai T Y, Li J N, Lewis F L. Data-
driven flotation industrial process operational optimal con-
m gÄzÆ
. , 2017, 43(6): 993−1006)
trol based on reinforcement learning. IEEE Transactions on
Industrial Informatics, 2018, 14(5): 1974−1989
ñ ² ÀÆ6§ónÜgÄz
I[:¢¿Æ¬ïÄ). 2016
¼
ÀÆnØó§a¬Æ
14 Jiang Y, Fan J L, Chai T Y, Lewis F L. Dual-rate opera-
tional optimal control for flotation industrial process with
unknown operational model. IEEE Transactions on Indus- . ÌïÄǑóL§$1,
trial Electronics, 2019, 66(6): 4587−4599
ä, g·AÄ5y, rzÆS.
15 Rojas D, Cipriano A. Model based predictive control of a E-mail: JY369356904@163.com
rougher flotation circuit considering grade estimation in in- (JIANG Yi Ph. D. candidate at the
termediate cells. Dyna, 2011, 78(166): 29−37 State Key Laboratory of Synthetical
770 g Ä z Æ ò 45
_ ÀÆ6§ónÜgÄz
I[:¢¿Æ¬ïÄ). ÌïÄ
Automation for Process Industries, Northeastern Univer-
ǑE,óL§nØ9Eâ.
sity. He received his master degree in control theory and
engineering from Northeastern University in 2016. His re-
search interest covers industrial process operational con- E-mail: jiayao neu@163.com
trol, networked control, adaptive dynamic programming, (JIA Yao Ph. D. candidate at the
and reinforcement learning.) State Key Laboratory of Syntheti-
cal Automation for Process Industries,
Northeastern University. His research
[å ÀÆ6§ónÜgÄz interest covers process control theory and technology for
I[:¢¿BÇ. 2011
¼ú complex industry process.)
ÆÆó§XÆ¬Æ (
{IUI{Zæ²áÆéÜ). Ì ÆUn ¥Ió§¬, ÀÆ
ïÄǑóL§$1, ó Ç, IEEE Fellow, IFAC Fellow. 1985
Daìä£Ä¬ä. © ¼ÀÆÆ¬Æ . ÌïÄ
Ï&ö. Ǒg·A, U)Í, 6§ó
E-mail: jlfan@mail.neu.edu.cn ngÄznØ!{Eâ.
(FAN Jia-Lu Associate professor at the State Key Lab- E-mail: tychai@mail.neu.edu.cn
oratory of Synthetical Automation for Process Industries, (CHAI Tian-You Academician of
Northeastern University. She received her Ph. D. degree Chinese Academy of Engineering, pro-
from Zhejiang University in 2011. She was a visiting scholar fessor at Northeastern University, IEEE Fellow, IFAC Fel-
with the Pennsylvania State University during 2009 ∼ 2010. low. He received his Ph. D. degree from Northeastern Uni-
Her research interest covers networked operational control, versity in 1985. His research interest covers adaptive con-
industrial wireless sensor networks, and mobile social net- trol, intelligent decoupling control, and integrated automa-
works. Corresponding author of this paper.) tion theory, method and technology of industrial process.)