Brazilian Journal

of Chemical ISSN 0104-6632

Printed in Brazil
Engineering www.abeq.org.br/bjche

Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
dx.doi.org/10.1590/0104-6632.20160334s20140163

SMART MONITORING AND DECISION

MAKING FOR REGULATING ANNULUS
BOTTOM HOLE PRESSURE WHILE
DRILLING OIL WELLS
M. P. Vega1*, F. B. Vieira1, L. D. Fernandes1, M. G. Freitas1,
E. Russano1 and A. L. Martins2
1
Departamento de Engenharia Química, Universidade Federal Rural do Rio de Janeiro,
BR 465, km 7, CEP: 23890-000, Seropédica - RJ, Brasil.
Phone/Fax: + (55) 21 3787-3742,
E-mail: vega@ufrrj.br
2
PETROBRAS S.A./CENPES, Av. Hum Quadra 07, Ilha do Fundão, Rio de Janeiro,
CEP: 21494-900, Rio de Janeiro - RJ, Brasil.

(Submitted: November 11, 2014 ; Revised: April 27, 2015 ; Accepted: June 28, 2015)

Abstract - Real time measurements and development of sensor technology are research issues
associated with robustness and safety during oil well drilling operations, making feasible the diagnosis
of problems and the development of a regulatory strategy. The major objective of this paper is to use an
experimental plant and also field data, collected from a basin operation, offshore Brazil, for
implementing smart monitoring and decision making, in order to assure drilling inside operational
window, despite the commonly observed disturbances that produce fluctuations in the well annulus
bottom hole pressure. Using real time measurements, the performance of a continuous automated
drilling unit is analyzed under a scenario of varying levels of rate of penetration; aiming pressure set
point tracking (inside the operational drilling window) and also rejecting kick, a phenomenon that
occurs when the annulus bottom hole pressure is inferior to the porous pressure, producing the migration
of reservoir fluids into the annulus region. Finally, an empirical model was built, using real experimental
data from offshore Brazil basins, enabling diagnosing and regulating a real drilling site by employing
classic and advanced control strategies.
Keywords: Sensor; Expert system; Hydraulic; Control; Operational window.

INTRODUCTION mud-pulse telemetry pressure-while-drilling (PWD)

gives the ECD (Equivalent Circulation Density)
A drilling process presents a rotating drill string values while the pump is on and the ESD (Equivalent
which is placed into the well. A drilling fluid is Static Density) measurement when the fluid circula-
pumped through the drill string and exits through the tion is interrupted. The ECD value is expected to
choke valve (Figure 1). There are many disturbances increase gradually with the increase of well depth.
that produce fluctuations in the annulus pressure dur- However, abrupt increases in ECD or pressure peaks
ing oil well drilling. When the well is drilled, there can indicate potential operational problems. In addi-
are hydrostatic pressure increases because of the tion, fluid influx from the reservoir changes the well
growth of the well length. A monitoring tool named flow rate and the rheological properties of the fluid

*To whom correspondence should be addressed

970 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

mixture inside the well. Finally, the pipe connection the swab effect due to the pipe connection procedure
procedure, which requires stopping and starting the or due to the withdrawal procedure of the drill pipe;
pump, produces severe fluctuations in the well flow the reduction of the level of the mud in the annulus
rates. The geometry of the well imposes a pressure region due to lost circulation problems or mal-
gradient along the length due to the presence of the function in the procedure of the withdrawal of the
drilling fluid. The pressure balance between the well drill pipe; improper monitoring of mud density re-
section and the reservoir is primordial for operation duction while drilling a gas formation (gas-cut mud).
and security purposes. If the pressure in the well is Concerning the drilling process, maximizing the
higher than the pore pressure of the reservoir, the rate of penetration (ROP) into the well reduces the
operation is called over-balanced drilling, which may drilling cost, but increase cuttings production. An
induce lost circulation problems, however; if the pres- increase of solids concentration might produce the
sure in the well is lower than the pore pressure of the formation of a bed of cuttings (horizontal drilling) or
reservoir, an under-balanced drilling mode is charac- increase the loading of cuttings (vertical drilling).
terized, allowing the migration of the reservoir fluids Friction losses and flow rate are related straightfor-
into the well annulus, if the formations are permea- ward and proportionally; however, a flow rate in-
ble. The over-balanced drilling is the most used crease produces hole cleaning, reducing solids con-
method for drilling oil wells, minimizing the risk of a centration. As a result, depending on the design of
blow-out, which produces the penetration of reser- the well, one may observe an inverse response for
voir fluid into the well until reaching the surface. the annulus bottom hole pressure, that is, as the flow
Most field cases, offshore Brazil, are drilled inside rate increases, the annulus bottom hole pressure ini-
an operational window, possessing as constraints: tially decays, due to the enhancement of hole clean-
pore pressure (minimum limit) and fracture pressure ing, reducing solids concentration. Then, the annulus
(maximum limit), which defines the mud density bottom hole pressure increases due to the increase of
range. friction losses. The rheology plays a complex role
concerning annulus bottom hole pressure, altering
hole cleaning, friction losses and potentially produc-
ing pressure overshoots after the circulation stops.
Considering dynamic conditions, a highly pseudo-
plastic nature is the desired behaviour for the drilling
mud.
In order to assure drilling inside the operational
window, there is a regulatory strategy for the annulus
bottom hole pressure, which contains compression
pressure, hydrostatic pressure, friction pressure loss,
differential pressure across the choke and atmos-
pheric pressure, despite process disturbances (Perez-
Téllez et al., 2004). Traditionally, in normal drilling
operations, the choke valve is adjusted manually. The
fluid composition and pressures are evaluated based
on steady-state values, and the choke valve is ad-
justed accordingly. The main problem of pressure
control during drilling is that there are no measure-
Figure 1: Oil well drilling structure. ments of pressure available during the periodic dis-
turbance, namely the pipe connection procedure,
The under-balanced drilling mode, through a per- when mud circulation stops. Wind et al. (2005) em-
meable formation, constitutes an operation mode ployed an electro-magnetic transmission system,
through which reservoir fluids (oil/gas) penetrate which might have problems due to the signal attenu-
continuously into the well, diluting and reducing the ation in deep wells. Monitoring data from the tools
mud pressure, configuring a kick scenario, which known as mud-pulse-telemetry pressure–while-drill-
might lead, if uncontrollable, to a gush at the well- ing (PWD) and mud logging are very efficient ways
head (blowout). Different situations may produce to implement process monitoring and control and
kick and blowout: hydrostatic pressure of the mud also to anticipate drilling problems (hole cleaning,
inferior to the pore pressure of the reservoir, for wellbore stability, fluid gelification, kick detection,
maximizing penetration of the drill bit; occurrence of breathing/ballooning, hydrates). In fact, the tools

Brazilian Journal of Chemical Engineering

 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 971

operate by sending monitoring data while the drill methodology developed. Finally, an empirical model
fluid is circulating. Reeves et al. (2005) developed a was built, according to the methodology presented
system which integrates, into the drill string, a signal by Vega et al. (2008), providing a confident model
cable, which, however, is disconnected during the based on neural networks for use in a real drilling
pipe connection procedure. Jenner et al. (2004) de- environment, employing classic and advanced con-
veloped a technique named continuous circulating trol schemes in the smart monitoring and decision
system (CCS), based on a mechanical device able to making program.
continuously pump the drill fluids, even during pipe
connections. Nygaard et al. (2006, 2011, 2013a,
2013b) implemented classic and predictive control- THE EXPERIMENTAL DRILLING UNIT
lers, through simulation studies, for under balanced
drilling of a gas-liquid phase system, using the choke The well drilling unit (Figure 2) was built using a
opening index as the manipulated variable in order to drill string of 6 m, containing in-line sensors of flow
control the annulus bottom hole pressure. A new ap- and density (Metroval - RHM20), based on the Cori-
proach for pressure control, while drilling, is named olis effect, and an in line pressure transducer (SMAR
Management Pressure Drilling - MPD. MPD creates - LD301-M). A mud pump (Weatherford - 6 HP), con-
a pressure profile for staying inside the operational nected to a frequency inverter (WEG), a choke de-
window (i.e., pore pressure and fracture pressure), vice (choke valve – ASCO - 290PD-25MM) and
controlling frictional losses and hydrostatic pressure, valves of the butterfly type (Bray – series30/31),
(Fossli & Sangesland, 2006). which are connected to the feeding tanks, are used to
Real time measurements, development of sensor manipulate the experimental unit (input variables).
technology, mathematical modelling, optimization All these input variables can change the annulus bot-
and control are tools that allow operating under the tom hole pressure, which is the output variable of the
desired pressure levels, being associated with robust- program for smart monitoring and decision making.
ness and safety of drilling operations, making feasi- The unit has two feeding tanks containing water (8
ble the diagnosis of spurious situations and acting for ppg) and mud (15 ppg – pseudoplastic behaviour),
regulation purposes. Thus, control and automation of making feasible injection of varying solid concentra-
drilling operations is a required activity for the future tions into the annulus through the butterfly valves
challenges of petroleum engineering, primordially (Bray – series30/31). This scenario represents the
under a scenario of narrow operational windows. A implementation of different rates of penetration. The
review indicates that most papers in the literature idea is to use water in order to simulate the drilling
deal with intelligent monitoring (Alvarado et al, fluids, and mud, to simulate the drill cuttings. This
2004; Mohaghegh, 2005; Nikravesh et al., 2002; configuration allows the implementation of different
Zhang, 2004; Sheremetov et al., 2005 and Shere- rates of penetration, without using a bit or solids in-
metov et al., 2008, Hermann, 2014) without linkage jection, since this would be a very difficult experi-
with regulatory control studies. Nowadays, an expert mental task. Concerning the control of the experi-
operator monitors the annulus bottom hole pressure mental unit, different values of the rate of penetra-
data and identifies undesirable events. tion can be implemented by varying the relative
The relevance of the present paper is the con- opening index of the butterfly valves, devices that
struction of an experimental unit, which presents the can also be employed as manipulated variables in
most important characteristics of the drilling process, order to control bottom hole pressure.
and using the experimental unit to validate smart The oil well drilling system is represented in the
monitoring and the decision making program, based structure of the experimental unit, which has an an-
on regulatory feedback control. As a result, the main nulus region, a pump, a choke valve and a drill bit,
objective of this paper is to perform automated drill- producing pseudosolids, experimentally implemented
ing, assuring an annulus bottom hole pressure inside through regulating the feeding of water/mud, using
the operational drilling window, employing smart the of butterfly-type valves. As a result, the solids
monitoring and decision making by selecting the ap- injection is made directly by employing the mud
propriate input variable (manipulated variable). Sce- tank. Different kinds of drilling phenomena can be
narios concerning the rejection of disturbances on captured in the experimental unit. The transient na-
the rate of penetration (ROP), set point tracking of ture of the annulus bottom hole pressure, due to the
the annulus pressure, inside the operational window, inherent phenomena of the growth of the length of
and the kick phenomenon are implemented experi- the well, the modifications in density and viscosity,
mentally for analysing of the performance of the affecting the hydrostatic pressure and frictional losses,

Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
972 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

can be implemented using the feeding tanks contain- Several operational parameters and disturbances
ing water and mud. In fact, by feeding of the unit may impact the annulus bottom hole pressure of the
using the water tank and, after promoting the in- well, such as rate of penetration, flow rate, rheology
creasing of the mud content, recycling the exit of the and the kick phenomenon. Experimental tests con-
unit to the water tank, makes feasible the study of cerning rejection of disturbance (rate of penetration),
increasing density levels. This scenario increases tracking of the annulus pressure set point under in-
both frictional losses and the hydrostatic pressure. creasing density levels, and kick disturbance are
Besides, using mud to feed the unit and then increas- analysed in this paper.
ing the water content through the butterfly valve ena- For regulation purposes concerning the annulus
bles a decrease of the density levels. This scenario re- bottom hole pressure, the fluid density and rate of
duces both frictional losses and the hydrostatic pres- penetration, actuated by modifying rheology and
sure. The pipe connection procedure can be imple- solid concentration in the annulus region, can be em-
mented experimentally through executing the stop- ployed. However, these manipulated variables do not
ping and the starting of the pump. Also, the kick or alter the annulus bottom hole pressure at the same
the lost circulation problem (mud loss) can also be velocity, due to the nature of the system to be distrib-
implemented experimentally, being detected through uted, which presents a large dead time for mud den-
an increase or decrease of the annulus flow levels, sity changes. The rate of penetration is improper for
respectively. regulation purposes when the pipe connection proce-
dure occurs, due to the requirement of stopping the
drill bit. In fact, concerning the rate of penetration,
the main objective during drilling is maximizing its
value, which reduces costs. Modifying the choke
opening index strongly alters the annulus bottom
hole pressure and can impact the intrinsic nature of
the drilling well, which may act as an open-closed
system. Using the mechanical apparatus reported by
Jenner et al. (2004), the mud flow is another in-
tended candidate for a manipulated variable, impact-
ing the annulus bottom hole pressure due to frictional
losses and also altering the residence time of the
mud, modifying well cleaning. As a result, depend-
ing on the smart monitoring diagnosing tool, which
classifies the specific scenario and provides the se-
verity of the problem during oil well drilling, the se-
lection of the manipulated variable is ROP, flow or
choke device, in this sequence, to the extent that the
Figure 2: Oil well drilling experimental unit: 1) pres- regulation problem becomes more infringing.
sure transducer; 2) on-line flow and density sensor;
3) helicoidally positive displacement pump; 4) feed Identification
tank (density - 8 ppg); 5) feed tank (density – 15 ppg);
6) choke valve; 7) butterfly valve; 8) recycle valves; In order to build an experimental regulation
9) stirrers. scheme for the annulus bottom hole pressure, a math-
ematical model for the drilling plant is required. The
development of a rigorous mathematical model may
METHODOLOGY not be feasible for complex processes involving a
large number of differential equations and unknown
Experimental Aspects parameters (physical and chemical properties). The
identification through a transfer function of low or-
A computational program, using C++ language, der (first-order plus dead time, estimating the steady-
was written in order to monitor, diagnose and regu- state gain k, the time constant τ and the dead time td)
late the drilling unit, employing as output variable has proved to be a useful tool and is the most popular
the annulus bottom hole pressure and, as input varia- framework for empirical model development for the
bles, the flow, the opening index of the choke valve purpose of classic controller synthesis. The methods of
and the indices of opening of the butterfly valves. reaction curve (Ziegler-Nichols, 1942) and Sundaresan
Brazilian Journal of Chemical Engineering
 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 973

& Krisnaswany (1977) were employed in order to A three-layer feed forward NN was built, using
identify the oil well drilling unit. hyperbolic tangent and linear activation functions for
Concerning nonlinear modelling, the approach the hidden and output layers, respectively. The input
using neural networks (NN) is a popular strategy for layer contained 28 neurons, i.e., the neural network
empirical model development, even though the esti- inputs contained present time data and past time in-
mate of the many parameters can often be regarded formation concerning: time, hole depth, depth, TVD
as a difficult problem to be solved (Mönnigmann et (true vertical depth), inclination, internal pressure,
al., 2002; Paladino et al., 2000). Vega et al. (2008) annulus pressure, block position, temperature, weight
developed investigations on systematic techniques for on bit, RPM, torque, stand pipe pressure and flow.
nonlinear model identification, using bifurcation dia- These data were employed to build the dynamic non
grams, the characterization of the amount and type of linear map, since the equivalent circulating density
process data required to build nonlinear empirical (ECD), the neural network output, depends on those
models with satisfactory predictive capability and the real time data provided by PWD and mud logging
selection of nonlinear model structures, which are (neural network inputs). In accordance with standard
capable of capturing a wide variety of behaviours. cross-validation procedures (Pollard et al., 1992), a
Concerning empirical modelling, the neural net- hidden layer with an optimal number of neurons
works were validated in terms of the traditional meth- (seven neurons) was selected. In order to build the
ods (Pollard et al., 1992; Sriniwas et al., 1995) and neural network model, two independent data sets
in terms of their complex static and dynamic behav- (training and validation sets), containing 5000 data
iour, using bifurcation and stability analysis. As ob- points each set, were employed. As discussed by Pol-
served through many examples (Vega et al., 2008) lard et al. (1992), the cross-validation training method
the use of traditional validation tests is not enough to minimizes the problem of over fitting.
guarantee the successful use of neural networks for The influence of the initial guesses of the parame-
monitoring and control purposes, because model and ters of the NN, during the training phase, on the re-
plant can present distinct behaviours. The compari- sulting dynamic behaviour was also investigated. It
son between the bifurcation diagrams of model and was observed that the particular bifurcation diagrams
plant, using the AUTO (Doedel, 2007), assures a obtained depend on the initial guesses. However, in
good performance criterion for nonlinear identifica- all cases, the bifurcation patterns observed were quali-
tion purposes. As a result, bifurcation techniques are tatively similar. Besides, the choice of the activation
used to allow the development of confident neural function of the neurons was also studied. Both
network models, based on experimental data pro- sigmoidal and hyperbolic tangent functions led to
vided by PWD and mud logging from offshore Bra- very similar bifurcation patterns.
zil basins. It is shown that the bifurcation and stabil-
ity analysis of the neural networks can be very help- Control
ful for appropriate development and implementation
of the empirical model in real time problems for di- The schemes implemented inside the control mod-
agnosis and disturbance rejection purposes. ule of the program of smart monitoring and decision
System stability analysis under parameter changes making were classic feedback and nonlinear model
is unveiled by bifurcation theory. Non linear pro- predictive control. Classic feedback control and non-
cesses might present multiple steady states, sustained linear model predictive control were synthesized in
oscillations and travelling waves under parametric order to regulate ECD, as the main objective is drill-
changes (Ray and Villa, 2000). The validation of the ing inside the operational window, that is, above po-
NNs, using the methodology of Vega et al. (2008), rous pressure and below fracture pressure. Ray
based on bifurcation and stability analyses, em- (1983) and Luyben (1990) pointed out that classic
ployed well-known continuation methods. For the feedback PID controllers (Seborg et al., 2011) have
neural network discrete model, the stability charac- been successfully employed under the majority of
teristics are determined by the eigenvalues of the process conditions. The review papers by Embiruçu
Jacobian matrix of the nonlinear map, relating past et al. (1996), Richalet et al. (1978), Garcia et al.
and present data with the future output of the pro- (1989), Qin and Badgwell (1996), Rodrigues and
cess. The steady states are stable if all eigenvalues of Odloak (2003), Karra et al. (2008), Dittmar et al.
the Jacobian matrix (Floquet multipliers) are inside (2012) and Beschi et al., (2014) referred to success-
the unitary circle; otherwise, if any of the eigenval- ful classic controllers implemented in complex sys-
ues is located outside the unitary circle, the steady- tems, representing real industrial processes. Predic-
state solution is unstable. tive control is a class of control algorithms in which

Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
974 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

a dynamic process model is used to predict and opti- constraints imposed by the model equations. This
mize system performance. The first predictive con- formulation generates an optimal controller in open
trol techniques were developed after 1970 driven by loop. The feedback is included by implementing the
the inability of classical controllers to meet the crite- manipulated inputs calculated for the present mo-
ria of increasingly demanding performances (Garcia ment, and then the prediction horizon is shifted one
et al., 1989, Huyck et al., 2014). step forward; the solving of the problem then uses
In this work, classic controller tuning approaches new measures of the process (Henson, 1998).
developed by Ziegler-Nichols (1942) and Cohen- Predictive control can deal explicitly with the
Coon (1953) were employed. Identification of the various changes that affect the process, as it incorpo-
plant along the entire operational range allows the rates its mathematical model, interconnections, and
use of the gain scheduling approach, providing a satis- physical and economic constraints. In the case of
factory control for each different point of system op- operation under different conditions, the process
eration. Thus, the linear control system becomes model must be able to handle the different regions of
more robust, compensating the nonlinearity of the operation.
process by employing distinct controller parameters In the formulation of predictive controllers, the
for each operational condition (Ramirez-Garduno tuning parameters are the control horizon, the predic-
and Lee, 2007). tion horizon and the weights matrix of the objective
The nonlinear predictive control (NMPC) is an function. For a fixed prediction horizon, a small con-
advanced control strategy based on a nonlinear trol horizon generates more conservative actions for
model. In recent years, the NMPC is gaining promi- the manipulated variables and slower responses to
nence in industry for being able to deal with nonline- the output variables; it is noteworthy that a greater
arities, employing phenomenological models or non- control horizon produces the opposite effect. Be-
linear empirical models and, simultaneously, solving cause that the control horizon is linearly related to
the control problem and the dynamic real-time opti- the number of decision variables, for the problem of
mization, satisfying constraints for state variables nonlinear programming, a greater control horizon
and manipulated variables. However, the NMPC leads to a big computational effort. Using larger pre-
meets resistance from part of the industry because of diction horizon, similar to the control horizon effect,
the difficulty of implementation. Apart from the dif- produces a more aggressive control and increases the
ficulty of modelling, there is also the complexity of computational effort required. The weights matrix
the algorithm itself. However, the benefits of apply- depends on the scale used in the problem under in-
ing this class of controllers for processes with large vestigation. Typically, they are diagonal matrices
nonlinearities have been widely reported in the lit- with positive elements. The magnitude of the diago-
erature (Manenti, 2011). The optimization routine nal elements depends on the scale used and the rela-
calculates the control actions (control horizon) and tive importance between the variables.
evaluates the response of the model along the predic- Meadows et al. (1997) defined the non-linear pro-
tion horizon. As the prediction horizon is greater gramming problem according to Eq. (1). The objec-
than the control horizon, the last control action is tive function J(K) should be minimized by manipu-
repeated until the calculated value of the prediction lating the optimization variable u, which is the vector
horizon is achieved. The formulation of the predic- of control actions. This objective function consists of
tive control is a problem of nonlinear programming two terms: the first minimizes the difference between
with nonlinear constraints (Meadows et al., 1997). the controlled variable and its set point, the second
For predictive control purposes, a sequence of minimizes the difference between the control actions
control actions is calculated to minimize an objective calculated at time k, k + 1, k + 2, k + 3,..., producing a
function that includes future values of the output smooth effect on the profile of the control actions.
variables, based on the model of the process. The There is also a strict restriction to Δu , which ensures
solution of the optimization problem is subject to maximum and minimum variations between each
constraints on the input and output variables and also sampling time.

P M

∑( ) (
y ( k + i k ) -ysp ( k + i k ) Q y ( k + i k ) -ysp ( k + i k ) + ) ∑ Δu ( k + i k )T RΔu ( k + i k )
T
min J ( k ) = (1)
u
i =1 i =0

Brazilian Journal of Chemical Engineering

 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 975

Constraints for i = k,", k + P + 1 : the methodology for diagnosing and regulating the
output variable (annulus bottom hole pressure or
x i+1 = f ( x i , u i ) ECD) in a typical drilling operational window, that is,
above porous pressure and below fracture pressure.
yi+1 = h ( x i+1 , u i+1 )

Δu max ≥ Δu ≥ Δu min

The neural network model, trained with real drill-

ing data from offshore Brazil basins, was employed
as the internal model of a model-based control strat-
egy (nonlinear predictive control). It is noteworthy
that detailed knowledge of the process is not neces-
sary for purposes of empirical modelling, which is
very attractive for complex industrial systems. In
addition, the non-linear problem to be solved in
closed loop, every sampling time, has complexity in
accordance with the structure of the non-linear em-
pirical model adopted. Su et al. (1997) stated that
NNs are the structures most frequently used for the
purpose of building empirical models, being appro-
priate for predicting process output, requiring small
CPU time.

RESULTS

Scenarios Discussion
Figure 3: Smart monitoring decision tree.
A methodology using real time monitoring was
built for diagnosing and implementing decision mak- For over-balanced drilling operation, the compu-
ing for both experimental and offshore drilling sites. tational program, using real time monitoring and es-
A computational program was developed in order to timating porous and fracture pressures, indicates, in
provide safe drilling inside the operational window. the case that the pressure level is inside the safe range,
Different scenarios can be analyzed and qualita- a message informing that the pressure is inside the
tive/quantitative actions are recommended in order to safe region, which constitutes the operational window.
regulate annulus bottom hole pressure and ECD. The If seepage is detected, the preferred input variable
output variable, annulus bottom hole pressure or is ROP. Experimental changes in ROP are imple-
ECD, is impacted by the solids travelling in the an- mented by modifying the relative opening index of
nulus region, which alters the hydrostatic pressure; the butterfly valves, altering the solids injection in
by solids forming a cuttings bed in high inclined sec- the annulus region. The changes in ROP are associ-
tions; by modifications in the rate of penetration dur- ated with slow actions, being appropriate for cases
ing oil well drilling; by the depth increase of the oil where the output variable is situated 20% below frac-
well; by pipe rotation and by changes in flow rate ture pressure and 20% above porous pressure.
and in rheology. Fast input variables, i.e., flow and choke opening
As a result, in order to avoid the lost circulation index are recommended under a partial loss scenario
problem, stuck pipe, barite sag, kicks, pressure peaks, and if the output variable is outside the range com-
and hole cleaning problems, a smart-monitoring com- prising 20% below fracture pressure and 20% above
putational program was developed in order to diag- porous pressure. If the safe range is not achieved, a
nose potential problems and indicate alternative ac- density change is suggested for the inlet drilling
tions, minimizing the disturbance effects and regulat- fluid.
ing the drilling process, assuring operation inside a Total loss situations are managed through recom-
safe envelope (operational window). Figure 3 presents mending the injection of loss circulating materials,
Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
976 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

such as mica or CaCO3. In the case of a more severe system to a step change in the input, then measure
scenario, squeezing a cement plug is advised. Finally, the output as a function of time (process reaction
critical measures include emergency casing. curve), and use this response to determine the control
The in-line measurements of pressure, flow and parameters. The reaction curve methodology was built
density are the monitoring tools for diagnosing the for representing the experimental unit of oil well
experimental drilling unit operation. The in-line pres- drilling by a transfer function model (first-order plus
sure sensor provides, in real time, the output variable dead-time), Table 1. The tuning parameters, Table 2,
(annulus bottom hole pressure), a primordial in- for different operational levels were calculated
formation for the feedback control loop, which is the through the tuning strategies of Ziegler-Nichols
configuration employed by the decision-making pro- (1942) and Cohen-Coon (1953). In order to identify
gram. The flow sensor can diagnose pipe connection the plant and select the tuning parameters, different
procedures, the kick phenomenon and lost circula- levels of frequency (30-60Hz) and choke opening
tion problems. The density sensor is able to detect indexes (25%-95%) were employed, comprising the
the phenomena of kick, barite sag, cleaning problems entire operational range of the mud pump and the
and disturbances in the rate of penetration. choke valve, as can be observed in Tables 1-2. For
The routine of decision making comprises a pro- identification purposes, the magnitude of the step
gram module containing proportional-integral con- disturbance, applied to the choke opening index and
trol and nonlinear model predictive control, allowing to the pump flow were: 95-25%, 95-35%, 95-55%
variation of the output variable inside a safe opera- and 15-30 Hz, 15-40 Hz, 15-50 Hz, 15-60 Hz, re-
tional range. In the decision-making module, set point spectively. As can be observed in Table 1, the magni-
tracking is imposed through the movements of the tude and the shape of the output variable (annulus
manipulated variable, according to the controller pa- bottom hole pressure) depend on the magnitude of
rameters, which feed the computational program de- the step disturbance implemented in the input varia-
veloped in C++ language. The experimental drilling ble, indicating that the experimental unit presents a
plant is remotely operated, using a sampling time of nonlinear behaviour. In fact, the time constant and
0.1 seconds. Depending on loss severity, different the steady state gain depend on the operational level.
input (manipulated) variables are selected. In fact, if In addition, when the choke opening index is de-
seepage is detected, the choice is to use ROP as the creased and the flow rate increases, the experimental
manipulated variable. A detection of a partial loss plant presents the more pronounced nonlinear re-
indicates the use, as input variables, of pump flow sponse. As a result, controller tuning parameters
and/or choke opening index. Under a scenario of to- were obtained for the entire range of process condi-
tal loss, changing of the inlet density of the drilling tions, in order to manage the nonlinearity of the sys-
fluid is advised, with injection of lost circulation ma- tem and implement classic PI control (Table 2), using
terials, squeezing cement plugs and performing emer- the gain scheduling approach.
gency casing. In order to analyse the performance of the smart
For the purpose of regulation in real time, nonlin- monitoring program and decision making using drill-
ear analysis, plant identification and controller pa- ing real data from offshore Brazil basins, the tools:
rameter estimation were performed for different oper- PWD and mud logging were employed. A three-layer
ational levels. Concerning the project of the classic feed-forward NN was built, using hyperbolic tangent
controller, which is inside the program module of and linear activation functions in the hidden and out-
decision making, proportional and integral actions put layers, respectively. The 28x7x1 architecture was
were selected. The proportional mode makes imme- used to build the dynamic map, since it is assumed
diate corrective action, as soon as the error is de- that the equivalent circulating density (ECD) de-
tected. The integral control action eliminates offset, pends on actual and past values of time, hole depth,
but generates fluctuations in the controlled variable, depth, TVD (true vertical depth), inclination, internal
reducing system stability. However, a limited amount pressure, annular pressure, block position, tempera-
of oscillation can be tolerated, as it is often associ- ture, weight on bit, RPM, torque, stand pipe pressure
ated with a faster response. Because the measured and flow. Besides, these values may be easily
variable of the drilling experimental unit has noise, evaluated in-line with PWD measurements and mud
the derivative action was not included in order to logging.
avoid noise amplification. There are several methods The smart monitoring program and decision mak-
for tuning a control loop, which involve the develop- ing contained a control module that regulates the
ment of some form of process model. The most com- process output (ECD), using the classic feedback
mon tuning method requires the subjecting of the controller or the nonlinear model predictive control-

Brazilian Journal of Chemical Engineering

 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 977

ler. The classic controller parameters were estimated Table 2: Controller tuning.
using Cohen-Coon (1953) and Ziegler-Nichols (1942)
methods (Table 2). In the formulation of predictive Ziegler-Nichols Cohen-Coon
controllers, the tuning parameters are the control hori- Choke Pump Kc Ti Kc Ti
zon (M=2) and the prediction horizon (P=4). A maxi- 1530 0.337 5.436 0.373 1.694
mum variation of ± 5 Hz between each sampling 1540 0.103 7.390 0.132 1.238
25%
time was employed in order to impose real limits of 1550 0.057 8.602 0.084 1.046
site operation, concerning mud pump constraints. 1560 0.062 8.806 0.088 1.154
1530 0.761 4.525 0.826 1.619
Table 1: Plant identification 1540 0.241 7.125 0.287 1.521
35%
1550 0.110 8.227 0.149 1.198
Choke Pump Delay Time Gain 1560 0.093 9.462 0.128 1.321
Constant 1530 0.868 4.736 0.958 1.510
1530 0.027 0.023 2.299 1540 0.367 6.705 0.433 1.502
45%
1540 0.037 0.012 2.871 1550 0.234 7.752 0.289 1.468
25% 1560 0.133 9.274 0.181 1.329
1550 0.043 0.008 3.063
1560 0.044 0.010 3.187 1530 1.692 3.724 1.807 1.517
1530 0.023 0.024 1.281 1540 0.590 6.058 0.682 1.495
55%
1540 0.036 0.017 1.782 1550 0.323 7.590 0.400 1.424
35% 1560 0.204 9.322 0.271 1.425
1550 0.041 0.011 2.141
1560 0.047 0.012 2.362 1530 2.001 3.644 2.161 1.353
1530 0.024 0.021 0.926 1540 0.970 5.300 1.093 1.491
65%
1540 0.034 0.017 1.259 1550 0.447 7.661 0.549 1.471
45% 1560 0.287 9.277 0.377 1.457
1550 0.039 0.015 1.524
1560 0.046 0.012 1.726 1530 2.670 3.492 2.858 1.397
1530 0.019 0.025 0.723 1540 1.045 5.540 1.198 1.420
75%
1540 0.030 0.018 0.908 1550 0.537 7.830 0.664 1.474
55% 1560 0.375 9.140 0.487 1.487
1550 0.038 0.015 1.087
1560 0.047 0.013 1.248 1530 2.216 3.898 2.430 1.297
1530 0.018 0.021 0.519 1540 1.011 5.704 1.193 1.281
95%
1540 0.026 0.019 0.678 1550 0.627 7.566 0.780 1.390
65% 1560 0.480 8.950 0.612 1.521
1550 0.038 0.015 0.815
1560 0.046 0.014 0.927 Ziegler-Nichols Cohen-Coon
1530 0.017 0.023 0.443 Pump Choke Kc Ti Kc Ti
1540 0.028 0.017 0.544 95-25 0.994 0.091 1.242 0.017
75%
1550 0.039 0.015 0.657 30 Hz 95-35 2.229 0.074 2.728 0.014
1560 0.046 0.014 0.749 95-55 3.072 0.061 3.654 0.013
1530 0.019 0.019 0.389 95-25 0.207 0.092 0.345 0.009
1540 0.029 0.015 0.458 40 Hz 95-35 0.922 0.067 1.164 0.012
95%
1550 0.038 0.014 0.542 95-55 0.985 0.060 1.300 0.009
1560 0.045 0.015 0.628 95-25 0.261 0.068 0.358 0.010
Pump Choke Delay Time Gain 50 Hz 95-35 0.366 0.072 0.520 0.009
Constant 95-55 0.779 0.053 0.982 0.009
95-25 0.027 0.010 0.336 95-25 0.039 0.069 0.115 0.004
0 Hz 95-35 0.022 0.009 0.167 60 Hz 95-35 0.104 0.063 0.216 0.005
95-55 0.018 0.009 0.143 95-55 0.403 0.058 0.548 0.008
95-25 0.028 0.004 0.600
40 Hz 95-35 0.020 0.007 0.344 Experiments
95-55 0.018 0.005 0.264
95-25 0.020 0.005 0.861 In order to illustrate the methodology of the smart
50 Hz 95-35 0.022 0.005 0.539 monitoring and the diagnosing tool, experimental
95-55 0.016 0.006 0.412 tests were performed without implementing the sug-
95-25 0.021 0.001 1.104 gested corrective actions that would regulate the an-
60 Hz 95-35 0.019 0.002 0.745 nulus bottom hole pressure of the experimental drill-
95-55 0.018 0.005 0.575
ing unit. As a result, evolutions of events beginning

Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
978 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

from seepage until attaining a total loss are observed. psi), characterizing an overbalanced drilling sce-
As illustrated in Figure 4a, the monitoring program nario, is the constraint imposed on the annulus bot-
indicates that the process is operating regularly, in- tom hole pressure.
side the operational window. Next, a disturbance is The aim of the regulatory control test is to ana-
detected, which produces an increase of annulus bot- lyse the performance of the closed loop to reject a
tom hole pressure; as a result, the program suggests disturbance, previously detected by the smart moni-
reducing the rate of penetration, Figure 4b. The cor- toring diagnosing tool, employing the manipulated
rective action was not implemented and a partial loss variable (ROP, pump flow or choke valve opening
was detected; as a result, the methodology indicates index). The regulatory control test (rejection of load
that the choke valve opening index must be increased disturbance) was implemented through introducing a
and/or the pump flow rate must be decreased (Figure perturbation in the rate of penetration, which was
4c). The last scenario presents the suggestions of the experimentally produced by changing the drilling
program concerning operation outside the safe pres- fluid density: water (8 ppg) to drilling mud (15 ppg).
sure envelope (above fracture pressure), indicating As a result, the use of a higher rate of penetration,
that fluid density must be reduced and lost circulat- detected through measuring the annulus density (Fig-
ing material should be introduced into the system; ure 5a), produced a higher solids concentration in the
moreover, some critical measurements may also be annulus region, which increases annulus pressure.
necessary, like for example squeezing a cement plug The smart monitoring program detected a partial loss
(Figure 4d). and the choke opening index was employed as the
Figures 5-7 illustrate the experimental drilling input variable for regulating purposes. As can be ob-
unit implementations of the smart monitoring and served in Figure 5b, after attaining steady state con-
decision making, under different scenarios, namely, ditions, inside the safe range of the operational win-
rejection of load disturbance (ROP perturbation), dow, an annulus bottom hole pressure increase was
drilling inside the operational window, the region in detected after 2 minutes of operation, which was pro-
between the pore pressure (minimum limit) and the duced by the higher rate of penetration adopted. The
fracture pressure (maximum limit) and the kick phe- smart monitoring program suggested manipulation of
nomenon. the choke valve device in order to minimize the dis-
An experimental drilling unit test was performed turbance, because a partial loss severity was detected.
in order to employ the tool for diagnosing and deci- As can be observed, the feedback control scheme
sion making implemented at the automated drilling successfully regulated the annulus bottom hole pres-
unit. An operational window, placed above the porous sure by increasing the choke valve opening index
pressure (20 psi) and below the fracture pressure (55 (Figure 5c).

a) b)

c) d)

Figure 4: Smart monitoring and the diagnosis tool: a) operation inside the operational pressure window, b) smart
monitoring seepage detection; c) partial loss smart monitoring detection; d) total loss smart monitoring detection:
fracture pressure; pore pressure; collapse pressure.

Brazilian Journal of Chemical Engineering

 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 979

Figure 5: Decision making implementation a) rate of penetration increase diagnostic through the smart
monitoring device; b) annulus pressure regulation (controlled variable): fracture pressure; pore
pressure; collapse pressure; c) choke opening index manipulation (manipulated variable).

Next, a variable operational window, with the po- increasing annulus density values. In fact, the ROP
rous pressure ranging from 3.5-35 psi and the frac- modifies the solids concentration and might produce
ture pressure ranging from 50-110 psi, under an over- the formation of a bed of cuttings (horizontal drilling
balanced drilling scenario, is the constraint of the – high angle) or increase the loading of cuttings (ver-
servo analysis, implemented in the drilling experi- tical drilling – low angle). During oil well drilling,
mental unit, whose aim is tracking the desired value the pore pressure (minimum limit) and the fracture
for the annulus bottom hole pressure. Simultane- pressure (maximum limit) are the limits that define
ously, disturbance rejection is analysed based on the the mud density range. As a result, the drilling fluid
tool of smart monitoring and diagnosis, which de- hydrostatic pressure needs to be higher than the pore
tects increasing density levels. This load disturbance pressure, in order to avoid formation fluid invasion
was implemented, experimentally, by increasing the into the well. Simultaneously, the drilling fluid hy-
index of opening of the butterfly valve connected to drostatic pressure needs to be smaller than the frac-
the mud tank (15 ppg) and, simultaneously, by de- ture pressure, avoiding formation damage. When the
creasing the index of opening of the butterfly valve smart monitoring program detected a partial loss, the
connected to the water tank (8 ppg). Figure 6 pre- choke valve device is employed as the manipulated
sents an experimental test concerning drilling inside variable, successfully tracking the desired value of
the operational window, under a scenario of increas- the annulus bottom hole pressure, despite the
ing values of the rate of penetration, which produce disturbances imposed by ROP variations.

Figure 6: Smart monitoring of ROP (density) and decision making for regulating the annulus pressure
(controlled variable) through the choke valve opening index (manipulated variable); fracture pressure;
pore pressure; water butterfly valve opening index; mud butterfly valve opening index.
Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
980 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

Concerning the kick experimental drilling unit environment. In accordance with standard cross-vali-
test (Figure 7), the smart diagnosis tool detects a dation procedures, a hidden layer with an optimal
flow increase due to reservoir fluid migration into number of neurons (7 neurons) was selected for de-
the annulus region. Initially, an influx from the reser- scribing the nonlinear map. A NN trained and vali-
voir, which is characterized as a liquid kick, is imple- dated with 5000 data points presented predictive
mented through water injection into the annulus re- capacity without over fitting, as discussed by Pollard
gion. Because the liquid injected and the mud in the et al. (1992). Figure 8 presents the neural network
annulus region present similar densities, the annulus validation test, implemented using offshore Brazil
bottom hole pressure remains almost unmodified. In basin data not employed in the training procedure.
fact, the kick disturbance slightly increases friction The bifurcation patterns were analyzed for the con-
loss, because the annulus flow is increased. Next, an tinuation parameters: hole depth, depth, TVD (true
influx from the reservoir, configured as a gas kick, is vertical depth), inclination, internal pressure, annular
implemented by injecting gas into the annulus re- pressure, block position, temperature, weight on bit,
gion. The program identifies the kick scenario and RPM, torque, stand pipe pressure and flow. Using
suggests the shutting procedure, that is, stop the mud the methodology developed by Vega et al. (2008),
pump, close the blow out preventer, open the choke stable behaviour was identified for the nonlinear
line and close the choke valve. It can be observed neural network map, built with drilling data from
that there is an increase of annulus bottom hole pres- offshore Brazil basins, i.e., all Floquet multipliers
sure inside the well, until reaching the formation (eigenvalues of the Jacobian matrix) were inside the
pressure. In fact, because the annulus bottom hole unitary circle, Figure 9.
pressure stabilizes at the value of the porous pres-
sure, additional influxes are avoided. The gas/liquid
mixture is circulated out of the well, using a reduced
circulation rate, maintaining constant the annulus
bottom hole pressure, by choke valve opening index
manipulation. After this stage, the well presents sin-
gle-phase behaviour, according to the driller’s
method procedure (Lyons & Plisga, 2005).
100

80 Figure 8: Neural network validation test.

Pressure, psi

60

40

20

00
0 5 10 15
Time, min

Kick scenario
Monitoring:
Annulus flow increase
Action:
Shutting procedure for tracking porous pressure
Mud circulation resumed Figure 9: Stability analysis of the NN.
Controlled: annulus pressure
Manipulated variable: choke opening index
Finally, the empirical model built with real ex-
Figure 7: Smart monitoring and decision making perimental data from offshore Brazil basins was em-
under a kick scenario; fracture pressure; ployed for diagnosing and regulating purposes
pore pressure. through classic and advanced control strategies. As
can be observed, the nonlinear model predictive con-
Next, field data, collected from a basin operation, troller performance is superior to the classic feed-
offshore Brazil, were employed for building a non- back controller under a scenario of ROP disturbance
linear neural network model for implementing smart rejection, Figure 10. The manipulated variable avail-
monitoring and decision making in a real drilling able at the drilling site for regulation purposes was
Brazilian Journal of Chemical Engineering
 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 981

pump flow, Figure 11. The performance of the con- a mud pump (Weatherford - 6 HP) and a choke
trol schemes (classic and nonlinear model predictive valve (ASCO - 290PD-25MM) are the candidates
control) was analysed subject to a step disturbance for being employed as input variables (manipulated
on ROP, configuring a regulatory test, Figure 12. As variables), for annulus bottom hole pressure regulat-
can be observed, the neural network model, trained ing purposes.
with real drilling data from the tools: PWD and Mud A smart monitoring program was built based on
logging, was successfully employed as the internal in-line sensors. The decision making tool employed
nonlinear model of the advanced control loop classic feedback control and NMPC structures for
configuration. regulating the process output.
In order to assure the drilling operation, using as
constraint a pressure inside the safe envelope, vari-
ous manipulated variables can be employed, depend-
ing on loss severity (seepage, partial loss or total
loss). A nonlinear analysis, plant identification and
parameter controller estimation were implemented,
including classic and advanced control techniques.
The smart monitoring program and decision making
was implemented in order to guarantee drilling in-
side the operational window and also to reject dis-
Figure 10: Controlled variable (ECD) fracture turbances (fluctuations in the rate of penetration).
pressure; pore pressure. A kick experimental test was implemented in the
drilling unit through injecting liquid and gas into the
annulus region. The smart monitoring device identi-
fied an annulus flow increase, followed by an annu-
lus bottom hole pressure increase. The shutting pro-
cedure was suggested by the program. After the an-
nulus pressure stabilization, additional influxes were
avoided, as indicated by the flow/density sensor in
the annulus region.
First-order plus dead-time transfer function
models were employed for a classic feedback control
Figure 11: Manipulated variable (flow). loop. A neural network model was identified using
real drilling data from the tools: Mud logging and
PWD. Besides, a nonlinear identification was per-
formed based on traditional validation procedures,
bifurcation and stability analysis. The neural network
model was employed as the internal model of an
advanced control scheme (NMPC) in order to regu-
late the ECD, using pump flow as the manipulated
variable. The nonlinear predictive controller pre-
sented faster disturbance rejection than the classic
feedback controller. Finally, the smart monitoring
Figure 12: ROP disturbance. and decision making program was also validated
using real data from an offshore Brazil basin.

CONCLUSIONS
NOMENCLATURE
An experimental unit was built for analyzing
recurrent scenarios that occur during the oil well KC controller gain
drilling process. The experimental plant contains KP process gain
sensors in-line: flow, density (Metroval - RHM20) M control horizon
and pressure transducer (SMAR - LD301-M), for P predictive horizon
disturbance detection purposes. Two butterfly valves Q weighting matrix of controlled variable
(Bray – series30/31), connected to the feeding tanks, R weighting matrix of manipulated variable

Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016
982 M. P. Vega, F. B. Vieira, L. D. Fernandes, M. G. Freitas, E. Russano and A. L. Martins

td dead time and Informatics, 8, 69-79 (2014).

Ti integral time Huyck, B., Brabanter, J., Moor, B., Van Impe, J. F.,
y output variable Logist, F., Online model predictive control of in-
ysp output variable set point dustrial processes using low level control hard-
Δu variation of the manipulated variable ware: A pilot-scale distillation column case study.
Δumax maximum variation of the manipulated Control Engineering Practice, 28, 34-48 (2014).
variable Jenner, J. W., Elkins, H. L., Springett, F., Lurie, P. G.,
Δumin minimum variation of the manipulated Wellings, J. S., The continuous circulations sys-
variable tems: An advance in constante pressure drilling.
In: SPE Anual Technical Conference and Exhibi-
tion, no. SPE 90702, Houston, TX, USA (2004).
REFERENCES Karra, S., Shaw, R., Patwardhan, S. C., Noronha, S.,
Adaptive model predictive control of multivaria-
Alvarado, M., Sheremetov, L., Cantú, F., Autono- ble time-varying systems. Industrial & Engi-
mous agents and computational intelligence: The neering Chemistry Research, 47(8), 2708-2720
future of AI applications for petroleum industry. (2008).
Expert Syst. Appl., 26(3-8) (2004). Luyben, W. L., Process Modeling, Simulation and
Beschi, M., Dormido, S., Sánchez, J., Visioli, A., Control for Chemical Engineers. McGraw-Hill,
Two degree-of-freedom design for a send-on- Singapore (1990).
delta sampling PI control strategy. Control Engi- Lyons, W. C. and Plisga, G. I., Standard Handbook of
neering Practice, 30, 55-66 (2014). Petroleum and Natural Gas Engineering. Elsevier,
Cohen, G. H. and Coon, G. A., Theoretical considera- USA (2005).
tions of retarded control. Transactions of the Manenti, F., Considerations on nonlinear model pre-
ASME, 827-834 (1953). dictive control techniques. Computers & Chemi-
Dittmar, R., Gill, S., Singh, H., Darby, M., Robust cal Engineering, 35(11), 2491-2509 (2011).
optimization-based multi-loop PID controller tun- Meadows, E. S., Rawlings, J. B., Scokaert, P. O. M.,
ing: A new tool and its industrial application. Con- Discrete-time stability with perturbations: Applica-
trol Engineering Practice, 20, 355-370 (2012). tion to model predictive control. Automatica,
Doedel, E. J., AUTO07: Software for Continuation 33(3), 463-470 (1997).
and Bifurcation Problems in Ordinary Differential Mohaghegh, S. D., Recent developments in applica-
Equations. Concordia University, Montreal (2007). tion of artificial intelligence in petroleum engi-
Embiruçu, M., Lima, E. L. and Pinto, J. C., A control neering. J. Pet. Technol., 57(4), 86-91 (2005).
survey of advanced of polymerization reactors. Mönnigmann, M., Marquardt, W., Bifurcation place-
Polym. Eng. Sci., 36, 433-447 (1996). ment of Hopf points for stabilization of equilibria.
Fossli, B. and Sangesland, S., Controlled mud-cap Proceedings of the 15th IFAC World Congress,
drilling for subsea applications: Well-control Barcelona, Spain (2002).
challenges in deep waters. SPE Drill & Compl21 Nikravesh, M., Aminzadeh, F., Zadeh, L., (Eds.),
(2): 133-140. SPE-91633-PA, USA (2006). Soft Computing and Intelligent Data Analysis in
Garcia, C. E., Prett, D. M., Morari, M., Model pre- Oil Exploration. Elsevier Science (2002).
dictive control: Theory and practice – A survey. Nygaard, G., Naevdal, G., Nonlinear model predic-
Automatica, 25(3), 335-348 (1989). tive control scheme for stabilizing annulus pres-
Garduno-Ramirez, R., Lee, K. Y., Fuzzy gain-sched- sure during oil well drilling. Journal of Process
uling PID + decoupling control for power plant Control, 16, 719-732 (2006).
wide-range operation. In: Preprints of the Interna- Nygaard, G., Zhou, J., Automatic model-based con-
tional Conference on Intelligent Systems Applica- trol scheme for stabilizing pressure during dual-
tions to Power Systems (ISAP2007), Kaohsiung, gradient drilling. Journal of Process Control,
Taiwan, 1-6 (2007). 21(8), 1138-1147 (2011).
Henson, M. A., Nonlinear model predictive control: Nygaard, G., Godhavn, J.-M., Aamo, O. M., Hauge
Current status and future directions. Computers E., A novel model-based scheme for kick and loss
and Chemical Engineering, 23, 187-202 (1998). mitigation during drilling. Journal of Process
Hermann, G., Application of Artificial Intelligence Control, 23(4), 463-472 (2013a).
Techniques in Monitoring Drilling Processes. Nygaard, G., Carlsen, L. A., Nikolaou, M., Evaluation
Advances in Soft Computing, Intelligent Robotics of control methods for drilling operations with
and Control – Topics in Intelligent Engineering unexpected gas influx. Journal of Process Control,

Brazilian Journal of Chemical Engineering

 Smart Monitoring and Decision Making for Regulating Annulus Bottom Hole Pressure While Drilling Oil Wells 983

23(3), 306-316 (2013b). Mourelle, M. Vallasco, A. Abraham, M. Koppen

Paladino, O., Ratto, M., Robust stability and sensi- (Eds.), Proceedings of the Fifth IEEE Interna-
tivity of real controlled CSTRs. Chemical Engi- tional Conference on Hybrid Intelligent Systems,
neering Science, 55, 321-330 (2000). Rio de Janeiro, Brasil, Nov. 6-9, IEEE, 92-97
Perez-Téllez, C., Smith, J. R., Edwards, J. K., Im- (2005).
proved bottomhole pressure control for underbal- Sheremetov, L., Batyrshin, I., Filatov, D., Martinez,
anced drilling operations. In: Proceedings for the J., Rodriguez, H., Fuzzy expert system for solv-
IADC/SPE Drilling Conference, no. SPE 87225, ing lost circulation problem. Applied Soft Com-
Dallas, TX, USA (2004). puting, 8, 14-29 (2008).
Pollard, J. F., Broussard, M. R., Garrison, D. B. and Sriniwas, G. R., Arkun, Y., Chien, I.-L., Ogunnaike,
San, K. Y., Process identification using neural B. A., Nonlinear identification and control of a
networks. Computers & Chemical Engineering, high-purity distillation column: A case study.
16(4), 253-270 (1992). Journal of Process Control, 5, 149-162 (1995).
Qin, J. S., Badgwell, T. A., An overview of industrial Su, H. T., McAvoy, T. J., Artificial Neural Network
model predictive control technology. Prepr. Proc. for Nonlinear Process Identification and Control.
Chem. Process Control Conf., CPC V, Tahoe City, In: Henson, M. A., Seborg, D. E., (Eds.), Nonlinear
CA (1996). Process Control. Prentice-Hall, Englewood Cliffs,
Ray, W. H., Multivariable process control - A survey. NJ (1997).
Computers & Chemical Engineering, 7(4), 367- Sundaresan, K. R. and Krishnaswany, P. R., Estima-
394 (1983). tion of time delay time constant parameters in
Ray, W. H., Villa, C. M., Nonlinear dynamics found time, frequency, and Laplace domais. Can. J.
in polymerization processes – a review. Chemical Chem. Eng., 56, 257 (1977).
Engineering Science, 55, 275-290 (2000). Vega, M. P., Lima, E. L., Pinto, J. C., Use of bifur-
Reeves, M. E., Payne, M. L., Ismayilov, A. G., Jelli- cation analysis for development of nonlinear
son, M. J., Intelligent drill string field trials demon- models for control Applications. Chemical Engi-
strate technology functionality. In: Proceedings neering Science, 63, 5129-5140 (2008).
for the SPE/IADC Drilling Conference, Amster- Wind, J., Weisbeck, D., Culen, M., Successful integra-
dam, The Netherlands (2005). tion of electromagnetic MWDLWD technology
Richalet, J., Rault, A., Testud, J. L., Papon, J., Model extends UBD operation envelope into severely de-
predictive heuristic control: Applications to indus- pleted fields. In: Proceedings for the SPE/IADC
trial processes. Automatica, 15, 387-402 (1978). Drilling Conference, no. SPE/IADC 92617,
Rodrigues, M. A., Odloak, D., MPC for stable linear Amsterdam, The Netherlands (2005).
systems with model uncertainty. Automatica, 39, Zhang, Z., Ch. Zhang (Eds.), Agent-Based Hybrid
4, 569-583 (2003). Intelligent Systems: An Agent-Based Framework
Seborg, D. E., Edgard, T. F., Mellichamp, D. A., for Complex Problem Solving. Lecture Notes in
Doyle, III F. J., Process Dynamics and Control. Artificial Intelligence 2938, Springer Verlag
3rd Edition, John Wiley & Sons (2011). (2004).
Sheremetov, L., Batyrshin, I., Martinez, J., Rodri- Ziegler, J. B. and Nichols, N. B., Optimum settings
guez, H., Filatov, D., Fuzzy expert system for for automatic controllers. ASME Transactions,
solving lost circulation problem. In: N. Nedjah, L. 64, 759-768 (1942).

Brazilian Journal of Chemical Engineering Vol. 33, No. 04, pp. 969 - 983, October - December, 2016