6th IFAC Conference on Management and Control of Production

and Logistics
The International Federation of Automatic Control
September 11-13, 2013. Fortaleza, Brazil

Modelling, analysis and control of

electroplating line modelled by P-time
event graphs.
T. Becha ∗ R. Kara ∗ S. Collart Dutilleul ∗∗ J.J. Loiseau ∗∗∗
∗
Laboratoire de Conception et Conduite des Systèmes de Production,
route de Hasnaoua BP 17, Université Mouloud Mammeri de
Tizi-Ouzou, 15000, Algérie (e-mail: redouk@ yahoo.fr)
∗∗
Institut Français des sciences et technologies des transports, de
l’aménagement et des réseaux, ESTAS/IFSTTAR, Villeneuve d’Ascq,
France, ([email protected] )
∗∗∗
LUNAM université, Institut de Recherche en Communication et
Cybernétique de Nantes, UMR-CNRS 6597, 1 rue de la Noë, 44321
Nantes cedex 03, France (e-mail:
[email protected])

Abstract: In this paper, we propose a method to deal with performance evaluation and control
of electroplating line modelled by P-time event graphs. The dater functions allow us to describe
the dynamical behaviour of a given P-time event graphs in the standard algebra by means of
implicit discrete system. From this formalism, different problems can be addressed. When we
assume that the behaviour of the modelled system is 1-periodic, a linear programming problem
can be derived from the discrete implicit system, to answer both, the minimum (maximum)
cycle time and the optimal initial sojourn time of tokens. In a control context, we show how the
discrete implicit system may be used to design a control law for a electroplating system using
the concept of positive invariance.

Keywords: Discrete event systems, P-time event graphs, Performance analysis, Constrained
control.

1. INTRODUCTION sider that the modelled system is operating initially which

generates an initial residence time of tokens.
A Petri net is a graphical and mathematical tool that is
On the other hand, the control of electroplating lines, well
useful for modelling, analysis, and control of discrete event
known under the name of Hoist Scheduling Problem or
systems. Particularly, when we consider systems involving
HSP introduces certain problem of control and analysis.
synchronization and parallelism phenomena and temporal
By considering a cyclic behaviour, two methods have been
constraints, the first step consist in providing a Petri net
introduced in scientific literature to deal with the HSP
model, the second step gives an algebraic model of the
problem. The first one provides a valid scheduling in an
considered system. Thus, many algebraic models have been
acceptable time ((Lei, 1991), (Caux, 1997), (Zou, 2003),
introduced in literature for these specifics class. In dioid
(Che A, 2004)), the second method we are interested in
algebra, the min plus (maxplus) dioids give a state space
takes into consideration the difficulty or the impossibility
representation of timed event graphs (Baccelli, 1992) that
to compute in real time a cyclic schedule. Hence, the
is useful for the performance analysis, resource optimiza-
approach consists in computing a valid off-line schedule
tion (Gaubert, 1995) and constrained control (Amari et al.
characterized by a time event graph, and to compute on-
2012). Declerck (2005) have proposed an interval model in
line the transitions firing dates of the obtained graph.
the dioid algebra for the P-time event graphs, this model
permits us to address some features as liveness, estimation The aim of this work is to model, analyse and control
and control. In the standard algebra, Guezzi (2010) have an electroplating line described by a P-time event graph
proposed the polyhedral model that is used to performance (Collart, 1997). Exploiting the state space representation
evaluation of timed and time event graphs. More recently, given in (Kara et al. 2013), we propose an equivalent model
(Kara et al. 2013) introduced a new model based on the that allows to take in consideration the initial residence
implicit discrete system representation. This representa- time of tokens in the Petri net model. When we assuming
tion had been used to constrained control of P-time event a 1-periodic behaviour, a linear programing problem is
graphs. In ((Guezzi, 2010), (Kara et al. 2013)) the authors derived which permits to get the optimal cycle time and
suppose implicitly that the initial sojourn time of tokens is initial residence time of tokens of the considered system,
equal to zero. In other terms, the initial residence time is this problem is tackled for the first time in this paper.
nul, which is an opposition with the hypothesis generally A feedback gain that gives the residence time of tokens at
assumed in the timed Petri nets. In this paper, we con-

978-3-902823-50-2/2013 © IFAC 311 10.3182/20130911-3-BR-3021.00027

IFAC MCPL 2013
September 11-13, 2013. Fortaleza, Brazil

each firing step is synthetized to an electroplating line that

permits us to change a cycle time (or rate production) and
to reach asymptotically to desired trajectory.
This paper is organized as follows: The main definitions
used in this paper are recalled in section 2. In section 3, Fig. 1. Part of P-time event graph
we give a linear programing that permits us to deal with
the problem of performance evaluation and compute the By associating the variable θj (k) as the date of the k th
optimal initial sojourn time. The control in closed loop is firing of a transition Tj ∈ T and o Tj the set of its input
applied to an industrial process in section 4. We conclude places. For each transition Tj , we can write:
in Section 5.
θj (k) = θi (k − m0i ) + wi (k), (1)
2. PRELIMINARIES where,
θi (k): the date of the k th firing of the transition Ti ∈ o Pi ;
In this section, we present some basic definitions that will m0i : the initial marking of the place Pi ∈ o Tj ;
be used in the sequel. wi (k): the sojourn time of tokens in the place Pi for the
Definition 1. An ordinary P-time Petri net is a six-tuple, k th firing;
N = (P, T, I, O, m0 , IS) where, The sojourn time must verify: ai ≤ wi (k) ≤ bi .
• P = {P1 , P2 , ..., Pn } is a finite set of n places represented From equations (1) the following matrix form can be
by circles; derived:
• T = {T1 , T2 , ..., Tm } is a finite set of m transitions
represented by bars; such that, P ∩ T = ∅ and P ∪ T 6= ∅; τ =l
• I : P × T → {0, 1}, is the input function such that
X
E · θ(k) = Aτ · θ(k − τ ) + w(k), (2)
I(Pi , Tj ) is called the value of the arc directed from Pi to τ =1
Tj ;
• O : P × T → {0, 1}, is the output function such that where, E ∈ Zn×m , Aτ ∈ Zn×m ,
O(Pi , Tj ) is called the value of the arc directed from Tj to w(k) ∈ Rn+ , l = max(moi ) and θj (k) = θj (0) ∀ k ≤ 0.
Pi ;
• m = [m1 , m2 ..., mn ]T ∈ Nn represents the marking
vector, mi is the marking of the place Pi , m0 is the initial To get a state-space representation, we must decompose
marking and N is the set of natural numbers; each place containing more than one token to several
• IS is the static interval function such that: Pi ∈ P → places and transitions where each place contains at most
IS(Pi ) = [ai , bi ] ∈ R+ × R+ ∪ {+∞}, where 0 ≤ ai ≤ bi is one token. By this operation, the model (2) can be written
the temporal interval of the place Pi . in the following generalized state-space representation:
E · x(k) = A · x(k − 1) + q(k), (3)
A place Pi (transition Tj ) is called an input place (tran- where, a ≤ q ≤ b.
sition) of a transition Tj (place Pi ) if and only if there
exists a directed arc from Pi to Tj (from Tj to Pi ). Place Remark 1. To obtain the model (3), the authors have
Pi (transition Tj ) is called an output place (transition) of supposed that the initial sojourn time of tokens in all
transition Tj (place Pi ) if and only if there exists a directed places of the net is equal to zeros. In fact, in some model,
arc from Tj to Pi (from Pi to Tj ). The following notations the electroplating line for instance ((Bonhomme, 2001),
will be used, o Tj is the set of input places of the transition (Collart, 1997)) the authors suppose that the system is
Tj , o Pi is the set of input transitions of the place Pi . initially operating, so that the initial sojourn time of
tokens may be non-zeros. It is then, necessary to operate
We call P-time Event Graphs a particular class of P-time some adaptation on the model that permits us to remain
Petri nets where each place has exactly one input and one in the situation given by (3).
output transition.
To take into account the initial sojourn time value, we
In a P-time Petri nets, a token must stay in a place Pi make some adaptations on the P-time event graph, and an
during the minimum residence duration ai . Before this equivalent model is obtained, the motivation is exposed on
duration, the token is unavailable to contribute to the an example.
firing of its output transition Tj ∈ Pio . The value bi ,
represent the maximum residence duration after which the
token must leave the place Pi (and can contribute to the 2.2 Example
enabling of the downstream transitions). If not, the system
falls into a token-dead state. So, the token is available to • Transformation 1:
fire the transition Tj in the time interval [ai , bi ]. We said Considered in dioid algebra to get the state space
that a P-time event graph has an admissible behaviour if representation, and in (Guezzi, 2010) to provide the
and only if his evolution do not lead to a token dead-states. polyhedral representation. This transformation does
not take into account explicitly the initial residence
time values of tokens.
2.1 The implicit systems model (Kara et al. 2013)
• Transformation 2:
Consider the P-time event graph given by the Fig.1.: The case considered in this work, the token in the

312
IFAC MCPL 2013
September 11-13, 2013. Fortaleza, Brazil

n̄i
X
ai ≤ qis ≤ bi .
s=1

Consider the dater functions associated to the transitions

added of the P-time event graph, by construction, we
obtain an implicit discrete model of the form (3), which
must verify the temporal constraints of the form:
a ≤ L · q ≤ b. (4)
Where a, b are defined below, q is the sojourn time vector
of the obtained P-time Petri net, and L is a matrix of size
Fig. 2. Techniques used to duplicate a place. N × n (with N is the number of the places of the obtained
graph).
place P2 has an initial sojourn time ∈ [a1i , b1i ] and Remark 2. By construction the model (3) is an implicit
the token in the place P3 has an initial sojourn time system, if E = In , the systems (3) is called an explicit
∈ [a1i + a2i , b1i + b2i ]. system.

3. PERFORMANCE EVALUATION
2.3 The equivalent model
Let us consider a 1-periodic behaviour given by the follow-
In the sequel, we expose the method used for the duplica- ing equation:
tion of the places initially marked with taking into account
the initial residence time of tokens.
x(k + 1) = x(k) + λ · u, (5)
where, x(k) is the dater vector, λ is the cycle time and u
is an M -vector with component are all equal to one.

By replacing the last equation in (3) we get,

(E − A) · x = q − λ · E · u,
which leads to

x
!
( −C −IN E · u ) · q = 0N +M +1 . (6)
λ
Fig. 3. Equivalent Petri net model.
subject to : a ≤ L · q ≤ b and x ≥ 0M .
Let us consider the part of P-time event graph given by the Where, N (M ) the number of the places (transitions) of
Fig.3., to get equivalent model, we decompose each place the equivalent event graph model.
Pi containing m0i tokens (m0i ≥ 1) to n̄i (n̄i = m0i + 1) The minimum (maximum) admissible cycle time, and the
places (Pis , s = 1, ..., n̄i ), with: optimal associated sojourn time of token can be computed
by solving the following linear programming problems.

n̄i : The number of intermediate places; 

msi : The marking of the intermediate place Pis ;  minimize (maximize) λ

 under constraints :

IS(Pis ) = [asi , bsi ] : the static temporal interval associated E · u · λ = C · x + q, (7)
to the intermediate place Pis , with 0 ≤ asi ≤ bsi ;
 a ≤ L · q ≤ b,

qis : The sojourn time associated to the intermediate


x ≥ 0M .
places;
n̄i − 1 : the number of intermediate transitions. 3.1 Application to an electroplating line

Such that, Let us consider the P-time event graph given by the
• msi = 0, for s = 1; Fig.4a. representing an electroplating line, see for more
Pn̄i
• m0i = s=1 ms ; details ((Bonhomme, 2001), (Collart, 1997)). To obtain the
Pn̄i s i Pn̄i s implicit state space form, we must build the model given
• ai = s=1 ai , bi = s=1 bi .
by Fig.4b. The dynamic behaviour of the P-time event
graph of the Fig.3b. can be expressed by the following
By this transformation, the temporal constraints on the state space form:
sojourn time of the place Pi given by ai ≤ qi ≤ bi will be E · x(k + 1) = A · x(k) + q(k + 1),
expressed as follows:
with:

313
IFAC MCPL 2013
September 11-13, 2013. Fortaleza, Brazil

 
1 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
 
0 0 0 1 0 0 0 0 0 0 0 1 1 0 0
 
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
 
0
and, L =  0 0 0 0 1 0 0 0 0 0 0 0 1 0.
0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 1 0 0 0 0 0 0 1
 
0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
 
0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 1 0 0 0 0

The LP problem (7) gives, the following cycle times:

λmin = 549 and λmax = 578, and the optimal sojourn
time:
qλmin = [40, 397, 58, 460.99, 58, 199.48, 58, 21.79, 66.83,
t
89.48, 124.68, 179.87, 274.31, 144, 32.20] and,
qλmax = [40, 400, 58, 466.1, 58, 200.90, 58, 33.71, 118, 112,
t
54, 176.01, 279.88, 149.09, 46.28] .
We have computed an optimal initial state in the sense
of sojourn time that gives a maximum (minimum) cycle
time.
Remark 3. With an initial condition on the state vector
x(1) and an input vector q, we can reach a desired
Fig. 4. Electroplating line. trajectory. In fact, this is an open loop control. It is well
known that the open loop control is sensitive to additive
−1 1 0 0 0 0 0 0 0 0 0 0
 
perturbations. This means that, the trajectory can diverge
 0 −1 1 0 0 0 0 0 0 0 0 0 from the desired one if the system is perturbed. To
 0
 0 −1 1 0 0 0 0 0 0 0 0 remediate to this drawback we investigate the possibility of
 0 0 0 0 1 0 0 0 0 0 0 0 control with a closed-loop law which provide the stability
 
 0
 0 0 0 −1 1 0 0 0 0 0 0 and the robustness towards time disturbances.
 0 0 0 0 0 0 1 0 0 0 0 0
 0 0 0 0 0 0 −1 1 0 0 0 0
 
4. CLOSED LOOP CONTROL OF
E= 1 0 0 0 0 0 0 0 0 0 0 0,
 
ELECTROPLATING LINE
 0
 0 1 0 0 −1 0 0 0 0 0 0
 0 0 0 0 1 0 0 −1 0 0 0 0
  4.1 Problem formulation
 0
 −1 0 0 0 0 1 0 0 0 0 0
 0 0 0 −1 0 0 0 0 1 0 0 0
 0

0 0 0 0 0 0 0 0 1 0 0
 Let us consider the P-time event graph of the electroplat-
 0 0 0 0 0 −1 0 0 0 0 1 0 ing line given bellow. We can see that the matrix E of
0 0 0 −1 0 0 0 0 0 0 0 1 the considered system is left invertible. Let E L be a left
inverse of E, then the unique solution of (3) is given by:
0 0 0 0 0 0 0 0 0 0 0 0 x(k + 1) = E L · A · x(k) + E L · q(k + 1), (8)
 
0 0 0 0 0 0 0 0 0 0 0 0 with, a ≤ L · q(k + 1) ≤ b.
0 0 0 0 0 0 0 0 0 0 0 0

0
 The system is actually linear (event) invariant with a
 0 0 0 0 0 0 0 0 1 0 0 constraints on the control vector.
0 0 0 0 0 0 0 0 0 0 0 0
 
0 0 0 0 0 0 0 0 0 0 1 0
0

0 0 0 0 0 0 0 0 0 0 0
 We want track an admissible trajectory z(k), k ∈ N (given
A = 0

0 0 0 0 0 0 0 0 0 0 1,
 by a 1-periodic behaviour)
0 0 0 0 0 0 0 0 0 0 0 0

0
 z(k + 1) = E L · A · z(k) + E L · qz .
0 0 0 0 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0 0 0 0 0
 Where qz is a vector of sojourn time that permits to get

0 0 0 0 0 0 0 0 0 0 0 0
 the trajectory z(k) with a certain cycle time.
0 0 0 0 0 0 0 0 1 0 0 0
 
0 Taking e(k) = z(k) − x(k), the system (8) is equivalent to
0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 e(k + 1) = E L · A · e(k) − E L · (q(k + 1) − qz ). (9)

and a ≤ L · q(k) ≤ b, with, One considers the class of systems described by the equa-
a = [40, 397, 58, 910, 58, 340, 58, 54, 54, 72, 54] ,
t tion (9), and looped by the control law formulated as
q(k) = [q1 (k), q2 (k), q3 (k), q4 (k), q5 (k), q6 (k), q7 (k), follows,
t
q8 (k), q9 (k), q10 (k), q11 (k), q12 (k), q13 (k), q14 (k), q15 (k)] ,
t
b = [40, 400, 58, 922, 58, 350, 58, +∞ + ∞, +∞, +∞] , q(k + 1) = F · e(k) + qz , (10)

314
IFAC MCPL 2013
September 11-13, 2013. Fortaleza, Brazil

Where F ∈ RN ×M is the feedback gains. for this, consider the P-time event graph of the Fig.4. and
The system (9) in closed loop form is given by a desired trajectory z(k) given by
e(k + 1) = Acl · e(k), (11) z(k + 1) = E L · A · z(k) + E L · qz ,
where Acl = E L · (A − F ).
where:
The state feedback control (10) is subject to the following E, A are defined above.
constraints: qz = [40, 400, 58, 300.47, 58, 128.37, 58, 14.94, 118, 112, 54,
t
a ≤ L · q(k + 1) ≤ b. (12) 256, 365.44, 221.63, 65.05] .
The constraints on the control variables generate a poly- We assume that z(1) = [113.19, 153.19, 553.19, 611.19,
t
hedral set P(S, ρ) defined by 377.2, 435.2, 207.19, 265.19, 787.21, 489.09, 584.28, 657.47] .
P(S, ρ) = {e(k), S · e(k) ≤ ρ} , The desired trajectory can be obtained, for instance, by
assuming a 1-periodic behaviour.
L·F ρ1 = b − qz
with S = and ρ = .
−L · F ρ2 = qz − a
The control must verify: a ≤ L · q(k) ≤ b, which leads to:
Now, Let P(P, p) be a polytopic convex set containing the P(S, ρ) = {e(k), S · e(k) ≤ ρ} ,
origin, defined by with S and ρ are defined as below.
P(P, p) = {e(k), P · e(k) ≤ p} ,
with P ∈ Rs×n , p ∈ Rs and p > 0. The constraints on the state are given by:
This set may represent, for instance, constraints on the −µ2 ≤ e(k) ≤ µ1 ,
state (dater) to be satisfied. with,
µ1 = z(1) and, µ2 = µ1
The last inequality can be reformulated under the form of
4.2 control by invariance the following polyhedral set:
P(P, p) = {e(k), P · e(k) ≤ p} ,
In the sequel, we recall some definitions and results about
the control by invariance of linear discrete time systems.

In µ1
where P = (with n = 12),and, p = .
Definition 2. (Blanchini,1999). A set P(P, p) is said to be −In µ2
positively invariant for the system (11) if and only if the
trajectory of the state remains in P(P, p) for any initial The linear programming problem returns the optimal con-
condition in P(P, p), i.e. vergence rate γ = 0.9, and the following associated matrix
e(k) ∈ P(P, p) =⇒ e(k + 1) ∈ P(P, p). (13) F:
Definition 3. (Blanchini, 1999). The polyhedron P(P, p)
0 0 0 0 0 0 0 0 0 0 0 0
 
is contractive for the system (11) if there is a positive γ < 1
such that if e(k) ∈ P(P, p) then e(k + 1) ∈ P(P, γ.p). 0 0 0 0 0 0 0 0 0 0 0 0 
0 0 0 0 0 0 0 0 0 0 0 0 
 
According to the results given in ((J-C Hennet, 1995),(Blan-  0 0 0 0 0 0 0 0 −0.05 0.92 0.08 0.03 
 
chini, 1999)), The system (11) is asymptotically stable if 0 0 0 0 0 0 0 0 0
 0 0 0 
and only if the following linear programming problem is

 0 0 0 0 0 0 0 0 0.01 0.02 0.87 0.01 
feasible. 0 0 0 0 0 0 0 0 0 0 0 0 
 
F =  0 0 0 0 0 0 0 0 0.02 0.02 0.04 0.97 .
 
 0 0 0 0 0 0 0 0 0.02 0.03 0.03 0.03 
minimize γ
  
  0 0 0 0 0 0 0 0 0.01 0.02 0.02 0.02 


 under constraints :  
 L L 0 0 0 0 0 0 0 0 0 0 0 0 

 H · P − P · E · F = P · E · A,  

 H · P − γ · p ≤ 0,

  0 0 0 0 0 0 0 0 −0.59 −0.56 −0.05 −0.03 
 0 0 0 0 0 0 0 0 0.65 −0.37 −0.03 −0.01 
 
D · P − D∗ · P = 0, (14)

 D · p ≤ 0,
 0 0 0 0 0 0 0 0 −0.01 −0.02 −0.87 −0.01 
0 0 0 0 0 0 0 0 −0.01 −0.01 −0.03 −0.96

D∗ · p ≤ 0,




 0 ≤ γ∗ < 1,


The state evolution versus number of firing and the control

D, D ≥ 0.
N ×M
input versus number of firing are depicted on Fig.5. and
Where A, E, P , p are defined above, F ∈ R , H ∈ Fig.6., respectively. It appears that the obtained feedback
R(2×M )×(2×M ) , D(D∗ ) ∈ R+(n×(2×M )) are matrices to be gains insures the asymptotic stability, the state and control
determined, and γ is the value of the convergence rate. constraints of the closed-loop system.
4.3 Numerical example 5. CONCLUSION

For an industrial necessity, we want to decelerate the Based on the modelling methodology proposed in (Kara
production of the system, our objective is then to design et al. 2013), we have dealt with the problem of perfor-
stabilizing feedback gains that permits us to change the mance analysis of a class of P-time event graphs. A linear
production rate, and guarantees for the defined system programming problem is formulated to compute, the min-
the respect of the control and state constraints. imum (maximum) cycle time, and an optimal initial state

315
IFAC MCPL 2013
September 11-13, 2013. Fortaleza, Brazil

REFERENCES
[1] Amari, S., Demongodin, I., Loiseau, J.J. & Martinez,
C. (2012). Max-Plus Control Design for Temporal
Constraints Meeting in Timed Event Graphs. IEEE
Trans on automatic control, 57(2), pp. 462-467.
[2] Baccelli, F., Cohen, G., Olsder, G.J., & Quadrat J.P.
(1992). Synchronization and linearity, An algebra for
Discrete Event Systems. Wiley, New York.
[3] Blanchini, F. (1999). Set invariance in control. Auto-
matica. 35(11). pp. 1747-1767.
[4] Bonhomme, P. (2001). Réseaux de Petri P-temporels,
contributions la commande robuste. Phd thesis.
Savoie university.
[5] Caux, C., & Pierreval, H. (1997). Solving a hoist
scheduling problem as a sequencing problem. Pro-
ceedings of MCPL’97 conference, Campinas (Brazil),
pp. 372-376.
[6] Che, A., & Chu, C. (2004). Single-track multi-hoist
scheduling problem: A collision-free resolution based
on a branch-and-bound approach. Int J Prod Res
42(12): 2435-2456.
[7] Collart Dutilleul, S., (1997). Commande Robuste
Dáteliers a contraites de temps de sejour: Apllication
a la galvanoplastie. PHD Thesis.
Fig. 5. State evolution [8] Declerck, P., & Didi Alaoui, M.K. (2005). Optimal
control synthesis interval descriptor systems applica-
tion to time stream event graphs.
[9] Gaubert, S. (1995). Resource Optimization and
(min,+) spectral theory. IEEE Trans. On Automatic
Control, vol. 40, pages 1931-1934.
[10] Guezzi, A. (2010). Modélisation, analyse de perfor-
mances et commandes des systèmes événements dis-
crets. Phd Thesis. Angers University.
[11] Hennet, J-C. (1995). Discrete-time Constrained Lin-
ear Systems in Control and Dynamic Systems, Vol.71,
C.T. Leondes Ed., Academic Press, 1995.
[12] Kara, R., Becha, T., Collart Dutilleul, S., & Loiseau,
J.J. (2013). An implicit system for modelling and
control of discrete event systems. 5th Symposium on
System Structure and Control Part of 2013 IFAC
Joint Conference SSSC, Grenoble, France, February
4-6,pp. 84-89.
[13] Khansa, W., Denat, J.P., & Collart Dutilleul, S.
(1996). P-time Petri nets for manufacturing systems,
Workshop on Discrete Event Systems (WODES’96),
Edinburg, Scotland, August 1996, pp. 94-102.
[14] Khansa, W., Aygalinc, P., & Denat, J.P.(1996). Struc-
tural analysis of P-time Petri nets, Conf. on Computa-
tional Engineering in Systems Applications (CESA’96
IMACS Multiconference), Lille, France, pp. 127-136.
[15] Lei, L., & Wang, T.J. (1991). The minimum common-
cycle algorithm for cyclic scheduling of two material
handling hoists with time window constraints. Man-
age Sci 37(12): 1629-1639.
[16] Zhou, Z., & Li, L. (2003). Single hoist cyclic schedul-
Fig. 6. Control inputs ing with multiple tanks: A material handling solution.
ComputOper Res 30: 811-819.
in terms of sojourn time of tokens. Using the concept of
control by positive invariance, a feedback gains matrix is
calculated, that permits to satisfy the constraints on the
state and control vectors. The both methods of control
(open loop control and closed loop control) are illustrated
on an electroplating line.

316