Appropriate control function

advertisement
HYBRID AVAILABILITY CONTROL MODEL FOR
VENTRICULAR ASSIST DEVICES
Diolino J. Santos Filho2, André V. V. Passanesi2, Vincent Donomai2, André C. M.
Cavalheiro2, Paulo E. Miyagi2, José R. Cardoso2, Aron J. P. Andrade1, Jeison Fonseca1
1
2
Departamento de Bioengenharia, Instituto Dante Pazzanese de Cardiologia, São Paulo (SP), Brasil
Departamento de Engenharia Mecatrônica e Sistemas Mecânicos, Escola Politécnica da USP, São Paulo
(SP), Brasil
E-mail: diolinos@usp.br
Abstract. Since an inappropriate behavior of Ventricular Assist Devices (VAD) may put a patient’s life at
risk, a hybrid availability control model for these devices is proposed in this paper. To reach this model,
a systematic approach is presented involving inductive and deductive reasoning which considers the
needs of each patient. Initially, a study of possible failure modes of this system was performed. Hence, a
Hazard and Operability Analysis (HAZOP) has been used along with a study of the characteristics of
failures and maintenance policies. Upon definition of the maintenance policy that effectively treats a
certain failure mode, the specific maintenance actions were described using the Production Flow Schema
(PFS) technique, a high-level form of Petri nets, leading to the construction of a logical model of
sequential activities that make up an availability control routine for the VAD. By successively refining the
PFS, the algorithm that represents the control routine can be generated and thus, automate the routine.
With a dynamic analysis of failure history, the formal models allow a continuous improvement of the
results acquired by the HAZOP study. This provokes an evolution of control models based on
maintenance models, involving either prevention or prediction, which are most adequate. The VAD
studied is in development at the Institute Dante Pazzanese of Cardiology.
Keywords: VAD, HAZOP, Maintenance, Petri nets.
1.
INTRODUCTION
Ventricular Assist Devices (VADs) are used to assist the pumping of blood in
patients with heart failure (Dinkhuysen, 2007; Andrade, 1999). The inappropriate
behavior of these devices can put a patient's life at risk (Bock, 2011). To better
understand this statement and the scope of this work it is important to emphasize the
following aspects:
 A VAD can have an advanced control system that provides it certain
features. The greater the complexity of this control system, the greater
the complexity of the control functions and refinement of feature that this
device will be able to perform.
 In this context, it is possible to create a control system that can anticipate
and treat failures during its operation by means of regenerative and
degenerative actions: the objective of the control system is to ensure
certain desired behavior of the system, admitting the implantation of
different solutions using different resources to act and intervene in the
device and in its way of interacting with the patient.
However, there are other important aspects that must be treated, e.g., the matter
of performance variations of these control system and of these devices in accordance
with their usage, considering component wear, behavior variations with time, or
changes in the circulatory system’s behavior that modify the predicted blood flow
conditions estimated during the project of the device.
The standard NBR 5462 defines failures as the termination of an item’s ability to
perform the required function, but the context of this work is in how to supervise the
operating state of a VAD, in other words, the existence of a control method that deals
with the question of maintaining the VAD available so that it can perform the required
functions is essential.
The European Federation of National Maintenance Societies (EFNMS) defines
maintenance as “the combination of all technical, administrative and managerial actions
during the life cycle of an item intended to retain it in, or restore it to, a state in which it
can perform the required function.” Thus, controlling the availability of a device is not
only the act of repairing something after it has entered a faulty state, but also the
monitoring and the prevention of the failure from occurring. Because of this, there are
several different availability control techniques which can be chosen for a system. The
selection of these techniques may be done based on the conventional models of
maintenance, presented in (Kelly, 2006):
 Corrective maintenance. Treats failures that are not periodic and that do
not have a detectable development time.
 Preventive maintenance. Prevents the effects of the failure in a system
and allows for a better management of the moments of intervention.
 Predictive maintenance. The failure modes have a detectable
development time, hence enabling their detection through predetermined
test routines.
Parting from this analysis, it is possible to conclude that the concept of
predictive maintenance can be associated with those of the classic preventive and
corrective maintenances (Jardine, 2005). Thus, the creation of an appropriate
availability control model whose purpose is to maintain the dynamic behavior presented
by a VAD working in the best condition for its usage is possible.
A partir desta análise, observa-se que o conceito de manutenção preditiva pode
ser associado aos conceitos de manutenção corretiva e preventiva clássicos (Desta
forma é possível gerar um modelo apropriado de controle de disponibilidade cujo
objetivo é manter o comportamento dinâmico apresentado por um DAV em
funcionamento na melhor condição para sua utilização. Consequentemente, são
destacadas questões pertinentes à necessidade de definição de procedimentos para testar
o comportamento do dispositivo, de modo a promover uma ação que leve o sistema ao
funcionamento desejado. Consequently, questions regarding the need of defining
procedures to test the behavior of the device, so as to promote an action that takes
system to the desired operation, arise.
Therefore, the objective of this paper is to propose an availability control system
for the usability of a VAD that is being developed at the Dante Pazzanese Institute of
Cardiology (Andrade, 1996; Bock, 2011), in order to maintain the dynamic behavior
presented by the system in the best condition for use, thus preserving the patient's life.
Section II of this paper presents the materials and methods used to develop the control
functions required. Section III presents the control functions and synthesized control
programs generated for the programming of the VAD’s controller. And section IV
outlines the conclusions of this work.
2.
MATERIAL AND METHODS
Figure 1 shows the methodology used to design an availability control system
for the VAD in study.
Risk Analysis and characterization of the identified
faults
Construction of the formal models of the availability
control functions associated to the faults
PN model refinement and analysis of the new models
generated
Transcription of the generated models to
implementable control algorithms
In Vitro tests of the algorithms in a test bench specific
for VAD simulation
Figure 1. Method for the synthesis of availability control functions of the VAD.
Initially the failure modes of the system were evaluated using the risk analysis
technique (HAZOP). This evaluation provides a better understanding of the behavior of
the system as a whole and of the failure mechanisms present in this, therefore enabling
an analysis of the nature of the failure modes collected. Thus, at this stage, the study of
the characteristics of failures for the adequate selection of the availability control
function is done.
After this study, it is possible to define the availability control actions that best
treats each failure mode. Once these actions have been determined, the model that
represent the availability control routine ideal for each failure are generated in the form
of a high-level Petri Net (PN).
Next, parting from the obtained models, the successive refinement of the highlevel PN is done to obtain a low level PN model, which is needed to obtain the control
program that will represent the availability control routine. To do this, it is necessary to
apply a method of model transcription, which will be presented later in this section, to
obtain a program, in ladder programming language, that can be implemented directly in
the controller of the VAD.
Once the low-level PN have been generated, it is possible to validate them
through computer simulations to guarantee that their basic performance is as desired.
The control programs are obtained from these validated models and are verifiable
through in vitro testing, so as to check the performance of the system in an environment
closer to its real working environment
The methods applied for the following steps will be presented next:
 Sort an assign the availability control functions.



Conduct the HAZOP-based risk analysis.
Model the control functions using PN.
Program the control functions in the VAD.
2.1 The Availability Control Functions
Considering the three main techniques found in the literature, which are the
corrective, preventive and predictive maintenance policies for a system, the availability
control functions can be sorted based on these policies. Thus we can classify these
control functions as:
a) Corrective availability control function. Corresponds to control actions
that must be performed when it is necessary to act upon the effects of a
fault, attempting to recover the performance of the device.
b) Preventive availability control function. Corresponds to the control
actions that are taken periodically as to avoid faults. For exemple,
considering that the clogging of the tube can occur as a function of the
VAD usage time, it is possible to schedule periodic control actions that
accelerate and decelerate the pump in a controlled manner seeking to
avoid the forming of clots.
c) Predictive availability control function. It is defined as the monitoring
and, based on the state system, the deduction of the moment of
intervention. It is noticeable that this availability control function is the
most complete, since it indicates the state in which the system is in at all
times, thus increasing its reliability. For this reason, this is the desired
control function when it comes to a system that requires high levels of
availability and safety, as is the case of this project, which deals with a
patient's health and quality of life. Nevertheless, this control function is
not suitable to be associated with all types of failure because it requires a
development time large enough for detection. Hence, a characterization
of the failure is required to select the adequate availability control
function.
Given the fact that the device in study is a mecatronic system, it has mechanical
and electronic parts and both of these are subject to fault. A mechanical system will
have, but not only, failure modes associated with wear; these have a detectable
development time and are not random, since they originate from a cumulative effect
caused by usage. So, with the checking of the state of the component, it is possible to
predict the moment in which it will enter a state of fault and intervene before this
happens.
On the other hand, electronic failures tend to have a different behavior. As an
example we have the burning of a sensor, an instantaneous and random fault, and
therefore undetectable. Thus, based on the characteristics of the fault, it is possible to
determine more appropriate associated availability control function. The characteristics
required for the usage of this method are the randomness of the failure and if it has
considerable development time (FDT) or not. Figure 2 illustrates the procedure of
allocation of availability control functions.
Select a Fault
Is it random?
N
Does it have
FDT?
Y
Does it have
FDT?
N
Y
N
Y
Define Corrective
Availability
Control Function.
Define Predictive
Availability
Control Function.
Define Preventive
Availability
Control Function.
Define Preventive /
Predictive
Availability Control
Function.
Figure 2. Availability control function assignment.
2.2 Risk Analysis
The study of Hazard and Operability (HAZOP), defined by the standard IEC
61882, is a risk-analysis technique that allows for a qualitative risk evaluation of the
possible faults of the system. Because it is a systematic and qualitative study, it enables
the assessment of systems that do not yet have a failure history. This analysis is divided
into three distinct parts, them being:
 Definition – consists in defining the scope and motives of the study.
 Preparation – the elements of the studied system as well as their specific
characteristics for the application are defined
 Evaluation - constitutes the determination of the deviations by applying
the technique and their analysis, checking if they are possible or not and
its causes and consequences
Since the HAZOP study used in this paper was conducted by a team during the
thematic project of which this work comes from, will give emphasis only to the
evaluation stage of this study. For this we defined two concepts needed for this step: the
element and the guide word. The element represents a division of the system which has
features to be evaluated and the guide word is defined as words or phrases that represent
a specific deviation from the expected behavior of the element or feature of the element
being measured. Examples of guide words are: NONE, LESS, MORE and REVERSE.
There are two possible ways of examining the stage of evaluation, namely: the
element-first and the guide word-first. For the element-first, we evaluate all deviations
generated by applying all the guide words to a given element before the evolution to the
next element. In the guide word-first, we evaluate the deviations generated by the
application of the same guide word to all the elements before changing the guide word.
The analysis of the deviations is done through a survey of the possible causes and
consequences of these deviations. The reason for the surveying of the consequences of
the causes and not of the deviations is given by the fact that the deviation is a
consequence in itself.
2.3 Modeling of Petri Nets
Dynamic systems may evolve from the occurrence of events, and are therefore
classified, from a theoretical standpoint, as Discrete Event Systems (DES) (Miyagi,
2007). Since faults can be interpreted as events that cause disturbance in the system, the
availability control routines are initialized because of these events and can therefore be
classified as a class of DES. Thus, to model the logic of the control functions, Petri Nets
can be used (REISIG, 1981).
Semantically, a Petri Net (PN) is a bipartite directed graph which has passive
and active elements connected by directed arcs that indicate the flow of information in
the system. There are several classes of Petri nets which can be used for modeling,
depending on the characteristics of the system in study (MURATA, 1989; REISIG,
1992; ROZEMBERG, 1998).
There are high-level nets that represent conceptual logic and sequencing of
activities, being able to model conflicts, parallelism and asynchrony. On the other hand,
there are marks nets of the type place/transition that are capable of accumulating
multiple marks in each place and can associate weights to the oriented arcs. To control
DES, interpreted nets like Condition/Event are commonly used; these are nets with
appropriate elements to represent external physical signs and are binary nets in which
their places cannot have more than one mark.
Thus, for the modeling of the maintenance routines concerning to this paper, an
initial use of High-Level Petri Nets is proposes. The models of these PNs can contain
natural or formal language associated with their elements- distributors, arcs and
activities - the latter being crucial, since an activity represents a macro-event. These nets
describe activities that may represent different events and states in macro-events, thus
facilitating the modeling of complex systems through the application of the concept of
successive refinement. So, it is possible to detail high-level nets in interpreted
Condition/Event (C/E) type nets.
From the moment in which we have a type C/E Petri net model, it is possible to
perform the transcription of these models into control programs. Therefore, in summary,
successive refinement is used to obtain a model interpreted PNs. As the low-level PNs
have the characteristics described earlier, it is possible to use a methodology that
transforms these models into control programs.
2.4 Cotrol Function Programs
Programmable Controllers (PLCs) are powerful tools in the area of automation
and control devices and have five programming languages standardized by the IEC
61131-3, one of them being the Ladder Diagram. This is a graphical language and
supports logical operations and complex functions (Miyagi, 2007).
Transcription of Petri Nets to algorithms Ladder is based on Mello et alii (2011).
The method is divided into four steps:
(1) determine which transitions are enabled;
(2) firing of enabled transitions, with the creation of marks in the output places;
(3) activation of outputs corresponding to the places with marks;
(4) initialization of system conditions, with the marks of the initial Petri Net.
Each element of the Petri Net (place and transition) has at least one variable
associated: each place has a variable with the suffix "Local" to indicate the presence of
a mark on it, and each transition has a variable with the suffix "Local" to indicate
whether it is enabled. In addition to these variables, each place can still have an output
variable associated with it, and each transition can have input variables associated with
it (corresponding to the enabling arcs and/or inhibitors).
3.
RESULTS AND DISCUSSION
Through the HAZOP analysis, 15 different failure modes of the system were
obtained. With the failure modes determined, the risk analysis was used to raise the
characteristics of each mode as to allow for an adequate selection of availability control
function. With these characteristics evaluated and using the method for allocation of
availability control functions illustrated in Figure 2, the results shown below were
obtained:
a)
Obstruction in the tube
Characteristics:
- Failure is random and originates from the human body;
- Does not have development time;
Appropriate control function:
- Corrective
b)
Disconnection of the tube by Random Reason
Characteristics:
- Random failure;
- Does not have development time because this failure is abrupt;
Appropriate control function:
- Corrective
c)
Motor Spikage due to Changes in Patient Behavior
Characteristics:
- Failure is random and originates from the human body;
- Does not have development time because of the spiking nature;
Appropriate control function:
- Corrective
d)
Sudden Uncoupling of the Motor
Characteristics:
- Failure is random and originates from the motor;
- Does not have development time;
Appropriate control function:
- Corrective
e)
Power Failure
Characteristics:
- Failure is random and originates from the electrical system;
- Does not have development time (characteristic of electrical
faults);
Appropriate control function:
- Corrective
f)
Inadequate Power Consumption by Electronic Circuit
Characteristics:
- Failure is random and originates from the electrical system;
- Does not have development time (characteristic of electrical
faults);
Appropriate control function:
- Corrective
g)
h)
i)
j)
k)
l)
m)
n)
Burned-out Motor
Characteristics:
- Failure is random and originates from the motor or the electrical
circuit;
- Does not have development time (characteristic of electrical
faults);
Appropriate control function:
- Corrective
Supervisory Controller does not act
Characteristics:
- Failure is random and originates from the control system;
- Does not have development time (characteristic of electrical
faults);
Appropriate control function:
- Corrective
No Current going through the Sensors
Characteristics:
- Failure is random and originates from the electrical system;
- Does not have development time (characteristic of electrical
faults);
Appropriate control function:
- Corrective
Disconnection of the Tube
Characteristics:
- Periodic failure caused by wear due to movement of the patient;
- Have development time;
Appropriate control function:
- Preventive Availability Control given the difficulty of checking
the connection state.
Sensors Not Calibrated
Characteristics:
- Periodic failure caused by normal use of the sensor;
- Has development time;
Appropriate control function:
- Predictive.
Motor Spikage due to Failure in the Sensors
Characteristics:
- Failure is random and originates from the electrical system;
- Has development time because the sensor failure is considered
the loss of calibration;
Appropriate control function:
- Predictive.
Low Speed of the VAD
Characteristics:
- Periodic failure caused by impeller or motor wear;
- Has development time;
Appropriate control function:
- Predictive.
Misalignment of Rotor Causing Magnetic Decoupling
Characteristics:
o)
- Periodic failure caused by impeller wear;
- Has development time;
Appropriate control function:
- Predictive.
End of Battery
Characteristics:
- Periodic failure caused by normal use of the VAD;
- Has development time;
Appropriate control function:
- Predictive.
Once the availability control functions were assigned for each failure, the next
step was to model the control functions themselves. The construction of the models was
performed by applying high-level PNs to represent the action to be taken in order to
formalize the semantic representation of the same. The maintenance routine should be
divided then into its active and passive elements, being that:
 The active elements correspond to the activities to be undertaken relevant
to the routine itself.
 The passive elements represent the intermediate states where decisions
that will define the next sequential actions to be performed are taken.
The availability control actions have models that are according to their
characteristics:
 For a corrective control function, the resulting PN will begin with the
discovery of the fault occurrence and, from that discovery, the realization
of this control function.
 The PN model of a preventive control function will have as its starting
action the end of the interval between interventions and the consequent
referral of patients to the hospital.
 A predictive control function will have activities to check the state of the
component that may fail to decide whether there is a need for change or
not.
As mentioned above, 15 different failure modes of the system were collected, 9
treatable by means of these corrective control functions, 2 treatable by means of these
preventive control functions e 4 by predictive control functions.
Figure 3 shows a high-level Petri net representing the availability control
function for the failure Cannula Obstruction, in order to illustrate the obtained results.
Figure 3. High-level PN model of availability Control Function for fault Cannula
Obstruction.
Following this, the synthesis of control programs is performed by applying
successive refinement of high level PNs to obtain the interpreted PNs that represent
them while keeping its semantic representation. It was then necessary to expand each
active element of the high-level PN and find their sub-states and actions.
So the macro-event is expanded into an input transition, their internal places
with internal transitions and an out transition. The internal states and transitions mold
the macro-event ensuring the expected behavior of activation of outputs. Then the
necessary control signals, either inputs or outputs, are placed for the operation of the
control routine.
Figure 4 illustrates the interpreted PN of the availability control function
obtained associated with the failure Obstruction in Cannula.
Figure 4. Interpreted PN model of availability Control Function for fault Cannula
Obstruction.
Finally, control programs associated with each of the availability control
functions modeled by means of interpreted PNs were generated. The methodology
involved four steps:
 Step 1: corresponds to transitions enabling. For each transition of the PN
their pre and post-conditions in the PN are identified, as described in
Figure 5. The resetting of the pre-conditions is performed at this stage to
ensure the desired dynamic behavior according to the model
implemented in a PN.
 Step 2: corresponds to the firing of transitions, ie, to update the local
state of each place and update the local state of each transition, as
illustrated in Figure 6.
 Step 3: corresponds to the activation of the output signals. For each of
the interpreted PN acts on the corresponding output, as shown in Figure
7.
 Step 4: corresponds to the initial marking of the PN. This rung is read
only in the first read cycle and can be seen in Figure 8.
Figure 5 Program for enabling the transitions.
Figure 6 Program for firing of transitions.
Figure 7 Program for output activation.
Figure 8 Program for marks initialization.
4.
CONCLUSIONS
This work proposes a new method for developing an availability control system
for VADs. The detailed causes of the failures described by HAZOP worksheet
facilitated the characterization of each failure, allowing for a quick assessment of the
best control function applicable. Based on the methodology for the allocation of
availability control functions proposed, it was possible to determine the control actions
for each failure in an objective and clear manner. Another interesting advantage found
in the study is that, as a technique for qualitative analysis, the study does not require a
history of failures, enabling the development of a control system that deals with risk
situations for devices in development, as is the case of this work.
The high-level PNs have shown themselves in being a strong tool for the
modeling of control procedures, since it allows for a high level description of the system
and a refinement in levels of detail in a systematic way. In this context there are two
effective contributions from the application of this approach: (i) because it is a graphical
model that accepts natural language it is an ideal tool for integrating multidisciplinary
teams facilitating the understanding without losing the formal description and semantics
and processes; (ii) with the use of PNs the automation of availability control processes
can be carried out, if applicable, through the use of methods for obtaining control
programs from the Interpreted PN.
Finally, it is important to emphasize that the methodology used for the
transcription of a PN models for control programs ensures the documentation of the
semantic aspects that are lost when generating a control program for low-level
language. Thus, maintenance of the code becomes feasible.
Therefore, this paper introduces a proposal of a model of availability control for
VADs which considers different types of control functions in order to evaluate the
system in its entirety. It is noteworthy that the large number of failures that can only be
treated by corrective availability control is due to the fact that the system is still under
development and also because part of this system is composed by a biological system,
introducing various random aspects in the study. However, for systems where the vital
parameters and their failure modes are unknown, the corrective functions enable the
conduction of a study of these features to continually improve the availability control
functions used. Thus, it is possible to improve the initially proposed model based on
data obtained from the system operating in patients.
AKNOWLEDGEMENTS
Thanks to Dante Pazzanese Institute of Cardiology (IDPC) for the workplace, to
the Research Foundation of the State of São Paulo (FAPESP) and to PET (Tutorial
Education Program MEC).
REFERENCES
1. Abdolrazaghi M., Navidbakhsh M., Hassani K. (2010), “Mathematical Modelling and Electrical
Analog Equivalent of the Human Cardiovascular System”, Cardiovasc Eng, v. 10, p. 45–51.
2. Alaydi, J.Y. (2008), “Mathematical Modeling for Pump Controlled System of Hydraulic Drive Unit of
Single Bucket Excavator Digging Mechanism”, Jordan Journal of Mechanical and Industrial Engineering,
v. 2, n. 3, p. 157-162. ISSN 1995-6665.
3. Andrade, A.J.P. (1998), “Projeto, Protótipo e Testes “In Vitro” e “In Vivo” de um Novo Modelo de
Coração Artificial Total (TAH) por Princípio Eletro-Mecânico de Funcionamento”, Tese de Doutorado,
UNICAMP. Campinas, SP.
4. Andrade, A.J.P. et al. (2008), “Mock circulatory system for the evaluation of left ventricular assist
device, endoluminal prothesis and vascular disease”, Artificial Organs, v. 32, p. 461-467.
5. Bell, R. (2005), “Introduction to IEC 61508”, Proceedings of ACS Workshop on Tools and Standards,
Sydney, Australi.
6. Bock, E.G.P. et al. (2008), “New Centrifugal Blood Pump With Dual Impeller and Double Pivot
Bearing System: Wear Evaluation in Bearing System, Performance Tests, and Preliminary Hemolysis
Tests”, Artificial Organs, v. 32, p. 329-333,
7. Bruciapaglia, A.H. et al. (2007), “Enciclopédia de automática: controle e automação”, Edgard
Blücher, v. 3.
8. Camp Sorrell, D. (2007), “Clinical Dilemmas: Vascular Access Devices”, Seminars in Oncology
Nursing, v. 23, n. 3, p. 232-239.
9. Cannon, C.P., Armani, A.M. (2007), “Preventive Cardiology: Insights Into the Prevention and
Treatment of Cardiovascular Disease”, 2. ed. Totowa, New Jersey: Humana Press, v. 1.
10. Cavalheiro, A.C.M. et al. (2011), “Specification of Supervisory Control Systems for Ventricular Assist
Devices”, Artificial Organs Journal, v. 35, n. 5, p. 465-470.
11. A.C.M. Cavalheiro et al. (2012), “Design of Supervisory Control System for Ventricular Assist
Device”. IFIP Advances in Iformation and Communication Technology, Lisboa, Portugal, p. 375-382.
12. Chang, Y., Gao, B., Gu, K. (2011), “A Model-Free Adaptive Control to a Blood Pump Based on
Heart Rate”, ASAIO Journal Adult Circulatory Support, p. 262-267.
13. Fonseca, J.W.G. et al. (2008), “A New Technique to Control Brushless Motor for Blood Pump
Application”, Artificial Organs, v. 32, p. 355-359.
14. Fonseca, J.W.G. et al. (2011), “Cardiovascular Simulator Improvement: Pressure Versus Volume
Loop Assessment”, Artificial Organs, v. 35, n. 5, p. 454–458.
15. Frank, P.M. (1992), “Principles of Model-Based Fault Detection”, In: Proceedings of International
Symposium on AI in Real-time Control, Delft, p. 363-370.
16. Fukamachi, K. et al. (1999), “Preoperative risk factors for right ventricular failure after implantable
left ventricular assist device insertion”, Ann Thorac Surg, v. 68, n. 6, p. 2181-2184.
17. Harewood, F., Grogan, J., Mchugh, P. (2010), “A Multiscale Approach To Failure Assessment In
Deployment For Cardiovascular Stents”, Journal of Multiscale Modelling, v. 2, n. 1 & 2, p. 21.
18. Hill, J.D., Reinhartz, O. (2006), “Clinical outcomes in pediatric patients implanted with Thoratec
Ventricular Assist Device”, Ped Card Surg Ann, v. 9, n. 1, p. 115-122.
17. IEC. (2006), “IEC 60812 – Analysis techniques for system reliability — Procedures for failure mode
and effectsanalysis (FMEA)”, IEC - International Electrotechnical Commission. [S.l.].
18. ISA. (1992), “ANSI/ISA-S5.1-1984 Instrumentation Symbols and Identification”, North Carolina.
19. ISO. (2007), “Medical devices — Application of risk management to medical devices - ISO14971”,
ISO - INTERNATIONAL STANDARD. [S.l.], p. 82.
20. Kurtoglu, T., Tumer, I.Y. (2008), “A Graph Based Fault Identification and Propagation Framework
for Functional Design of Complex Systems”, Journal of Mechanical Design. DOI: 10.1115/1.2885181
21. Lee, G.B., Zandong, H., Lee, J.S. (2004), “Automatic generation of ladder diagram with control Petri
net”, Journal of Intelligent Manufacturing, v. 15, p. 245-252.
22. Legendre, D.E.A. (2009), “In Vitro Comparative Analysis Between In Series and In Parallel
Cannulations for Ventricular Assist Device”, ASAIO Journal, v. 55, n. 2, p. 172.
23. Li, Z., Zhou, M. (2008), “Control of elementary and dependent siphons in Petri nets and their
application, IEEE Transactions on Systems, Man and Cybernetics, Part A: Systems and Humans”, vol 38,
n. 1, pp. 133–148, ISSN 1083.4427.
24. Park, J.W. et al. (2009), “Estimation of Native Cardiac Output of Patients Under Ventricular Assist
Device Support Using Frequency Analysis of Arterial Pressure Waveform”, Artificial Organs - Fifth
International Conference on Pediatric Mechanical Circulatory Support Systems and Pediatric
Cardiopulmonary Perfusion - Abstracts, v. 33, n. 5.
25. Reed, T.R., Reed, N.E., Fritzson, P. (2004), “Heart sound analysis for symptom detection and
computer-aided diagnosis”, Simulation Modelling Practice and Theory, v. 12, p. 129-146. ISSN 1569190.
26. Sales, T.P., Cavalheiro, A.C.M., Santos Filho, D.J. (2010), “Mathematical Modelling of The Human
Cardiovascular System”, 18° SIICUSP, São Paulo, SP.
27. Villani, E., Miyagi, P.E., Valette, R. (2006), “Modelling and Analysis of Hybrid Supervisory Systems:
A Petri Net Approach”, 1. ed. [S.l.]: Springer, v. 1.
28. Vural, K.M. (2008), “Ventricular assist device applications”, Ankara, Turkey: AVES Yayincilik Ltd.
29. Wilson, S.R. et al. (2009), “Ventricular Assist Devices: The Challenges of Outpatient Management”,
Journal of the American College of Cardiology, New York, New York.
30. Yi, W. (2007), “Physiological Control of Rotary Left Ventricular Assist Device”, Proceedings of the
26th Chinese Control Conference, Zhangjiajie, Hunan, China, p. 469-474.
Download