Automation of the Harvard Apparatus with these modifications.

The deployed upgrade allows for com-

plete automation of the Harvard Apparatus 1423 pulsatile blood
Pulsatile Blood Pump pump, with the capability to run sequences of conditions without
the need for manual intervention. [DOI: 10.1115/1.4007637]
Charles E. Taylor
Department of Biomedical Engineering,
1 Introduction
Virginia Commonwealth University,
401 West Main Street RM1229, 1.1 Background. The Harvard Apparatus (Holliston, MA)
Richmond, VA 23220 pulsatile blood pump (model 1423) is a piston pump capable of
single ventricle simulation in cardiovascular research. This model
e-mail: [email protected] is capable of reproducing left ventricular output similar to humans
and provides manipulation of specific pumping parameters. The
Zachary W. Dziczkowski positive displacement design and integration of check valves in
Department of Computer Science, the pump head creates a close approximation to the left ventricle
Virginia Commonwealth University, pumping mechanism. The ability to independently manipulate the
401 West Main Street RM4225, heart rate, stroke volume, and percent systole positions this design
Richmond, VA 23220 as a favorable means of constructing flow conditions within a
mock circulatory loop. This product’s established usage in ven-
e-mail: [email protected] tricular simulations was a deciding factor in the choice to include
this pump in the mock circulatory loop being developed by this
Gerald E. Miller laboratory [1–3]. It has been proven through these previous works
Department of Biomedical Engineering, to be a suitable reference standard for a pulsatile pump source in
Virginia Commonwealth University, the course of evaluating cardiac assist technologies.
401 West Main Street RM1229, The design developed by Harvard Apparatus involves the use
Richmond, VA 23220 of manual controls mounted on the case of the pump to control the
performance parameters (Fig. 1). This design requires the operator
e-mail: [email protected] to manually adjust each parameter at the pump, with no ability for
remote operation. The heart rate and percent systole settings are
assumed to be the operating conditions of the device after the
Producing accurate physiological circulatory conditions in vitro appropriate correction factors are applied from the manual [4].
is integral to the evaluation of cardiac assist technologies. The There is no ability to obtain confirmation and performance read-
ability to simulate cardiac function, normal or pathological, is back from the device.
dependent on the capabilities of the pump deployed for this pur-
pose. Presented is a reference standard for this in vitro analysis,
with automation features targeted for robust bench-top testing.
Cardiac performance is typically described in terms of stroke vol-
ume, heart rate, and percent systole. Respectively, these three set-
tings prescribe the volume of fluid ejected, the rate of pumping,
and the percentage of the pumping cycle spent in ejection. A pump
that provides settings for each of these parameters and precise
repeatability allows for accurate construction of simulation
conditions. These capabilities are present in the commercially
available Harvard Apparatus 1423 pulsatile blood pump. Modifi-
cations have been made to this particular model that allow for the
automation of its function and real-time performance determina-
tion. Discussed in this publication is the design and performance
of a modified 1423 pump that employs universal serial bus (USB)
communication in the control of its stroke volume, heart rate, and
percent systole. The percent systole is denoted as the phase ratio
on the hardware. Utilization of an embedded microcontroller
(MCU) allows for not only the digital communication via a com-
puter terminal, but process control of the subsystems maintaining
each parameter. Care was taken to preserve the mechanical
design employed by Harvard Apparatus; the modifications were
not invasive to the mechanical driveline of the pump. Electrome-
R
chanical design characterization was performed in SimulinkV
using the following SimscapeTM block sets: SimscapeTM Founda-
tion Library, SimElectronicsV, and SimMechanicsTM. This pro-
R Fig. 1 Isometric view of the Modified Harvard Apparatus Model
1423 pulsatile blood pump. This perspective on the device
vided an accurate model of the systems during the design process,
is able to show all the modifications made to the assembly:
which assisted in the deployment of the process controllers fastener connecting stroke volume actuator to lead screw (1),
with minimal prototype construction. Communication with the manual control knobs for heart rate and percent systole (2),
MCU is achieved with American Standard Code for Information switch for selecting manual knob or communication-based set-
Interchange (ASCII) commands delivered through a LabVIEW VI tings control (3), BNC output connector for logical indication of
interface. Continuous readbacks on fill/ejection rate, pump rate pumping phase (4), USB connection for communication with
(HR), percent systole (PS), and stroke volume (SV) are possible the device (5), and bootloader connection for updating the firm-
ware for the device (below USB connection). All the electronics
are mounted on the black side panel of this modified pump; the
Manuscript received January 6, 2012; final manuscript received August 7, 2012; only components not included on this panel are in the stroke
published online October 15, 2012. Assoc. Editor: Paul A. Iaizzo. volume actuation assembly and motor shaft encoder.

Journal of Medical Devices Copyright V

C 2012 by ASME DECEMBER 2012, Vol. 6 / 045002-1

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Table 1 Settings adjustment and features available on factory
model and automated version of the Harvard Apparatus pulsa-
tile blood pump

Setting Factory model Automated model

Heart rate
Actuation On unit dial Remote control via USB
Set point control Open loop Feed-forward PI control
Percent systole
Actuation On unit dial Remote control via USB
Set point control Closed loop Feed-forward PI control
proportional
control
Stroke volume
Actuation Hand crank Remote control via USB
Set point control Set screw Closed loop proportional
control
Setting data logging Not available USB communication-based

Fig. 2 A model of the drivetrain, designed in SolidWorks, and

later imported into SimulinkTM via the SimMechanicsTM Link
tool. The arrows indicate the movement of the respective link- scribed that the heart rate, percent systole, and stroke volume
ages. The elements of the drivetrain are the (1) cam connected must be changeable from a LabVIEW (National Instruments Cor-
to the driveshaft of the motor, (2) swing arm, (3) third linkage poration, Austin, TX) interface with real-time determination of
acting as a class 1 lever, (4) fulcrum of lever, and (5) translating the pumps performance for each of those parameters. The result-
linkage that attached to the hydraulic piston. ing updates to the pump are detailed in Table 1, a comparison
between the functionality delivered by the factory model and the
automated version. These improvements allow for appropriate
The supplied control mechanisms for the heart rate and percent integration into automated circulatory simulation systems, precise
systole are potentiometers (Fig. 1), which are interfaced to the condition generation, and accurate pump performance recording.
electrical board controlling the motor that actuates the hydraulic
piston. The generalized electrical schematic that applies to the
model 1423 pump is identical to that of the model 665 Intermedi- 1.2 PWM Motor Driving. Pulse width modulation (PWM)
ate Ventilator produced by Harvard Apparatus; this schematic is control is an efficient means of actuating direct current (DC)
presented in the ventilator’s user’s manual [5]. What is presented motors. The method necessitates the use of a steady DC voltage
in this ventilator schematic is a connection to a “cam switch,” an and a fast switching mechanism to achieve different motor speeds.
element that detects when the pump is filling (diastole) or ejecting The switching produces a square wave of voltage, by which the
(systole). This switch provides a mechanical feedback on the motor is driven. The duty cycle of the wave is proportional to
position of the piston, which in turn changes the speed in that the speed achieved by the motor. Semiconductors, preferably
particular phase to meet the settings for percent systole and heart MOSFET technology, are the means for executing the KHz to
rate. MHz switching frequencies needed for this method. The method-
The piston of the pump is actuated by translational motion ology for PWM control and semiconductor choice is the topic of
achieved by converting the rotational motion of an electric motion actuation research, with an abundance of resources avail-
motor’s driveshaft by means of a bar linkage, as depicted in Fig. 2 able [7]. The method employed in the motor control applications
[6]. The third link is a class 1 lever with a sliding collar as the ful- for this design was a PWM H-bridge, along with gate driver and
crum; this changes the mechanical advantage with the position of optoisolator integrated circuits (ICs). Use of an H-bridge allows
the piston. The fulcrum for the third link can also be translated the motor to be actuated bidirectionally from a single-side power
along the longitudinal axis of the link. This is the mechanism for supply and also allows for active braking; the optoisolator and gate
changing the magnitude of displacement in the fourth link per rev- driver, respectively, prevent loading effects and provide the current
olution of the first link, affecting the stroke volume in the piston. boost necessary to switch high-power MOSFETs in the H-bridge.
This translation of the fulcrum is achieved via a worm gear, with
a hand crank mounted on the top of the device (Fig. 1). A position
1.3 Shunt-Wound Motors. The motor supplied in the Har-
indicator and equivalent stroke volume scale are included on the
vard Apparatus pulsatile blood pump model 1423 modified in this
side of the device. The fourth link is constrained by roller bearings
publication is a Bodine Electric Company (Northfield, IL) NSH-
to only allow translation along the longitudinal axis and is inter-
34RH shunt-wound motor. From the information on the plate
faced to the hydraulic piston. The piston is constrained in the cas-
R attached to the motor, it is seen that this motor is capable of
ing by two O-rings with Christo-LubeV (Lubrication Technology
86 rpm at 29 lbf-in of torque while being driven with the specified
Inc., Franklin Furnace, OH) applied to reduce friction. This
115VDC. The horsepower rating is 1/15 and the current drawn at
mechanical driveline provides for an angular torque change and,
peak load is 1 A. The gearbox has a ratio of 1:20. The motor
with the friction of the piston, can provide significant motor
mechanical properties of inertia and damping could not be deter-
loading.
mined through available product literature.
Previous modifications to this pump have been focused on
retrofitting the hydraulic pump head. Two previous studies inves-
tigated the use of distensible membranes to create a physiological 1.4 SimulinkTM Physical Modeling. Construction of an
fluid-structure relationship in various forms [2,3]. Replacement of accurate computational model allows for tuning of process con-
the ball check valves with prosthetic valves for the purpose of trollers and simulating performance to identify operating points
flow field visualization has also been investigated [1]. This present for desired conditions. The SimscapeTM (The MathWorks, Natick,
study encompasses the design modification to automate a com- MA) toolboxes for the SimulinkTM (The MathWorks, Natick,
mercially available pump (Harvard Apparatus Model 1423) to MA) programming environment provide a means of accurately
achieve a remotely operated design capable of producing accurate modeling multidomain physical systems [8]. Elements for electri-
ventricular pumping profiles. The modification requirements pre- cal components, mechanical bodies, and hydraulic parts are

045002-2 / Vol. 6, DECEMBER 2012 Transactions of the ASME

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 3 Schematic of the run-time process control code developed for the Harvard Model 1423 design updates. The 200 Hz loop
processes the cam timing events, which serve to change the rpm settings, determination of heart rate, and calculation of per-
cent systole. This timing loop is also responsible for the PI controller with feed-forward (FF) section that dictates the pump
motor speed. The capture interrupt processes the optical encoder data and determines the speed. A slower 10-Hz timing loop
processes the positioning of the stroke volume position and resulting actuation. The communication section is placed in an
interrupt section to immediately process incoming (Rx) data. The low and high designations pertain to the priority of that inter-
rupt; the ranking within a low/high designation is available in the device manual. The pins utilized by particular functions are
designated next to each trace line. The blocks external to the run-time code section illustrate the systems this MCU is interfaced
to, with dashed lines representing assemblies.

Journal of Medical Devices DECEMBER 2012, Vol. 6 / 045002-3

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 available, with variables that are specific to the definition of the ing each phase of pumping for a particular heart rate and percent
part in its particular domain. This allows for data sheets and refer- systole setting is determined by the following equations:
ence material on parts included in the assembly to be used in con-
BPM 180 deg BPM
structing a numerical model highly reflective of the physical RPMsystole ¼ ¼ (1)
component. Interfacing elements that bridge between domains are ð% SystoleÞ 360 deg 2ð% SystoleÞ
also included, such as electromechanical motors and hydrome-
BPM
chanical pistons. ) RPMdiastole ¼ (2)
In the case of custom assemblies, reverse engineering purposes, 2ð1  % SystoleÞ
or for identifying parameters of components with no reference
information, there is the capability within the SimulinkTM envi- Using the equations for equivalent rpm in each phase, the maxi-
ronment to perform a parameter estimation task. Experimental mum rpm at a particular setting can be determined. rpm values
data from the physical system is used to fit unknown variables in above the rated speed of the motor elucidates the origin of the cor-
the computational model through regression analysis. This pro- rection factors in the user’s manual for the Model 1423 pertaining
vides the capability of constructing a high-fidelity computational to heart rate corrections for percent systole settings [4]. This
model from assemblies with unknown parameters or finely tuning design update includes a higher driving voltage of 145VDC,
known variables within the bounds of their error values. allowing for a broader range of operation than the base Harvard
Apparatus design. A higher voltage was not tested, as the
2 Materials and Methods 145VDC represented 126% of the rating for the motor, which was
as far out of specification as was deemed safe. The upgraded
2.1 Microcontroller. A microcontroller (MCU) was chosen design is capable of reaching a maximum speed of 105 rpm, which
as the process control hardware for the embedded design due to precludes simultaneously high heart rate and highly asymmetrical
its low cost and scalability. The Microchip Technology Inc. systolic ratios, but provides an operational range that was deemed
(Chandler, AZ) PIC18F2550 provided the features that were sufficient for most applications.
required in the design of the Model 1423 automation [9]. The The torque load on the driveshaft of the motor has a wide range
MCU has a native capture, compare, and PWM (CCP) module of variability, which necessitates a control design that can provide
that was used for control of the pump motor [10]. The capture the disturbance rejection to maintain the requisite rpm. A closed
function of the CCP module was used to register the transition loop PI controller with a feed-forward section was implemented
timing from an optical encoder attached to the pump motor drive- (Fig. 5). This controller topology allows for low volatility set
shaft for the purpose of rpm measurement. The MCU also point maintenance with the PI section and fast adaptation with the
included onboard analog to digital converters (ADC), which were feed-forward section. The feed-forward section takes advantage of
used to monitor position potentiometers and electrical board sup- the system’s excellent open loop control potential, but needs the
ply voltages. The key element to the controller was the universal set point maintenance after the rpm change. At an rpm set point
asynchronous receiver/transmitter (UART) module. This interface change, the PI controller is reset to prevent overshoot from roll-
provides the means of digital communication with the device, the over error.
primary objective in the automation of the Model 1423. A UART The error signal feeding the PI controller is taken from the rpm
to USB converter was implemented to enhance the connectivity of determination using the CCP module’s capture of the optical
the instrument to computer platforms, as RS-232 connections are encoder data on the driveshaft subtracted from the rpm setting
not common on modern hardware. Multiple timer modules for that pumping phase (Fig. 3). The optical encoder provides a
allowed for synchronizing events and sampling of data during the quadrature-based data sequence created by the optical wheel
runtime program, which provide crucial stability to data process- motion through the sensors. In this application, only one channel
ing and allow for consistently timed readback updates.

2.2 Pump Motor Control. Speed control of the pump motor

is enacted through an H-bridge PWM drive design. This method
was chosen for its precision in motor speed actuation, compatibil-
ity with the MCU motor driving functionality, and ability to
provide active braking. The H-bridge product employed was an
R
ADVANCED Motion ControlsV (Camarillo, CA) 25A20DD
driven by the CCP module on the PIC18F2550. A shunt regulator
was employed with the power supply to prevent overcharging of
the motor driver board supply voltage. Although the 25A20DD
includes optoisolation and gate drivers, the PIC18F2550 was not
capable of sourcing the current needed to drive the device. The
50 mA supply current requirement to drive the optoisolators on
the 25A20DD necessitated the use of low-side gate drivers
between the MCU and the motor driver board. This implementa-
tion serves to provide adequate wattage for the application and
provides an additional level of isolation protection for the low
voltage MCU.
The motor has two operating conditions: steady rpm for 50%
systolic ratio and switching rpm for non-50% systolic ratios set-
tings. In the case of the non-50% systolic ratios, the motor must
operate at a particular rpm for systole and another for diastole to Fig. 4 The computer-aided design (CAD) model of the stroke
produce differing duration times for filling and ejection. Each turn volume actuator at an isometric view showing the top of the as-
sembly. The gears are presented as cylinders with radial size
of the driveshaft on the pump motor corresponds to a fill/eject
equivalent to the pitch diameter. The interface of the actuator to
cycle on the piston; revolutions of the driveshaft per minute are the stroke volume lead screw (1). The back pane in this drawing
equivalent to simulated beats per minute (BPM). Time in systole (2) has been truncated to show only the section used for mount-
and diastole is determined purely by the speed of the motor in that ing the stroke volume actuator. The full back pane supports the
phase, as the mechanical linkage constrains half the revolution drivetrain shown in Fig. 2. The elements of actuator motor as-
(180 deg) to each phase. The angular velocity determination dur- sembly are a mounting spacer (3), gear box (4), and DC motor (5).

045002-4 / Vol. 6, DECEMBER 2012 Transactions of the ASME

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 5 The SimulinkTM model of the Harvard Apparatus model 1423 pulsatile blood pump. The subsystems that comprise the
larger model are presented with masks to illustrate the physical assembly they represent. Operating parameters are set on the
left side of the model: stroke volume (SV), heart rate (HR), percent systole (PS), and the power switch ON/OFF logic. The rpm
determination block uses Eqs. (1) and (2) to produce the required rpm settings for each phase of pumping. The “SV to
PivotPosition” lookup table in the upper left is an application of Eq. (3), providing the setting position from the SV value. The
cam switch block actuates the rpm switch to enact the correct rpm setting. The PI controller operates on the error of the setpoint
delivered from the preceding rpm switch and the driveshaft speed readback from the rotary encoder. The controller reset (lower
left of PI block) is enacted by a change in the cam switch state, allowing the feed-forward control from the rpm switch block to
dominate the speed change in the motor. The model has no outputs, as the hydraulic system this is connected to would need to
be modeled. The pump head block represents the translation of the piston and the friction pertaining to that motion.

is being monitored, due to the unidirectional motion. The sensor the lead screw assembly employed by Harvard Apparatus. The
has 1000 counts per revolution (CPR), allowing for adequate time design of the stroke volume actuator sought to reduce invasive-
resolution of speed change. The capture module interrupts at ness to the existing structure and preserve the functionality pres-
every 4th pulse; a faster capture rate caused too many interrupt ent. It was also desired that the actuator and position sensors be
events at high rpms that interfered with the MCU processing. This contained within the case for esthetics and splash guarding. The
capture rate equates to a 1.44-degree resolution of movement on actuator design was chosen to be a geared DC motor with a gear
the driveshaft. This precise rpm feedback with a high refresh rate train driving the post of the lead screw. Due to the internally
enables the PI controller to react appropriately to torque changes, mounted design feature, the interference of the assembly with the
providing excellent disturbance rejection. pumps linkage driveline and space constraints with the external
The feed-forward section of the controller is driven by the cam case produced the cantilever mounting of the motor (Fig. 4). The
switch and uses the rpm equations stated to predetermine what the installation of the motor, gear train, and bracket was achieved
rpm setting should be for that pumping phase. This set-point by removing the pump-mounting pane from the case. No modifi-
is added to the PI controller output, and this sum is scaled to the cations were made to the linkage driveline; only clearance holes
10-bit duty cycle input to the PWM module. Feed-forward were drilled through the mounting pane. The DC motor chosen
topology was implemented due to the failure of classic propor- was a high-torque (15 lbf-in), low-speed (60 rpm) design, as the
tional-integral-derivative (PID) architecture, even with gain sched- measurement of torque needed to drive the stroke volume change
uling, in accomplishing the rapid set point with minimal overshoot. could not be measured. A gear train was chosen as the easiest
This design takes advantage of the excellent open loop gain that means of working within the small clearance space provided. This
this system exhibits, but adding the needed closed loop control to design ensured that the torque of the motor would be adequately
maintain the set points during changing shaft torque loads. transmitted to the post. The ratio of the gear train is 1:1, due to the
The equivalent BPM and percent systole of the pump is moni- strength of the motor. The size of the gears is reflective of
tored by a timer module that tracks the transitions of the switch the distance from where there was room for motor in the case to
actuated by the cam mounted to the driveshaft of the motor. The the position of the lead screw post.
rising edge of the switch pulse signals the start of systole, with the
fall indicating the return to diastole. The following rising edge
signals the beginning of the next systolic phase and the conclusion
of the previous cycle. The capture of these three times provides
the values for calculations of BPM and percent systole. These val-
ues for heart rate and percent systole are those returned when the
MCU is queried for performance values.
This pump motor control design and sensor data created a plat-
form capable of executing the precise shaft speeds needed to
obtain the heart rate and percent systole performance needed. The
confluence of updated motor driver board electronics and PI con-
troller with feed-forward design produced a highly responsive
electromechanical system capable of maintaining the set points
through varying load conditions. This strong disturbance rejection
and fast set point adaptation is critical to achieving the perform- Fig. 6 The SimulinkTM model of the shunt wound motor, using
ance needed in this application, which was found to come only SimscapeTM elements, represents the electromechanical sys-
with the combination of the stated methods and hardware. tem of the motor accurately. An important factor in constructing
a shunt wound motor model is the mutual inductance between
the armature and the field coils, which is represented in the
2.3 Stroke Volume Control. The stroke volume automation upper left. The mechanical elements presented are consistent
necessitated the addition of an electromechanical actuator to drive with a DC motor model attached to a gearbox.

Journal of Medical Devices DECEMBER 2012, Vol. 6 / 045002-5

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 Fig. 7 SimulinkTM SimMechanicsTM equivalent of the SolidWorks model presented in Fig. 2. The light gray elements in this
model were created and connected automatically from the SimMechanicsTM Link. These blocks represent the rigid bodies and
constraints equivalent to the SolidWorks parts and assembly; the rigid bodies can be identified as those containing the center
of gravity symbol. The black elements were added to interface the SimMechanicsTM blocks with the model and provide an initial
condition for the geometry. The “rot” connection on the left of the figure is interface of the motor driveshaft to the linkage, and
the “trans” connector on the right of the figure is where the hydraulic piston is connected. The joint actuator for the pivot point
controlling the stroke volume (SV) necessitates the inputs of control parameter.

Stroke volume setting determination is achieved via a linear limit. If the error is outside the prescribed limit, the motor moves
potentiometer connected to the pivot point of the driveline link- in the appropriate direction until the error is within the limit.
age. This potentiometer is monitored by one of the ADC channels Communication update of the set point changes the error limit to
on the MCU, with position updates every 100 ms (Fig. 3). This zero, initiating movement. Upon a condition of zero error in the
update rate was chosen based on the determined speed of the position to set point, the motor is stopped and the error limit is
stroke volume actuator and resolution needed for this information. widened. This error limit window technique is in response to
The movement logic of the stroke volume is such that, upon slight movement in the pivot point during pumping and the lack
startup, the current position becomes the set point to eliminate ini- of electronics shielding in the prototype design, leading to spuri-
tialization movement. The movement of the stroke volume actua- ous ADC values.
tor is monitored by a 10-Hz check on the value of the current
position against the set point, with a test condition against an error
2.4 Computational Model. The computational model of
the design was utilized in the design process for both algorithm
development and design validation (Fig. 5). SimscapeTM tool-
boxes were used extensively in the construction of the three
physical domains present in this device: electrical, mechanical,

Fig. 8 The mock circulatory loop used to evaluate the perform-

ance of the upgraded design. Beginning at the modified
Harvard Pump (1), the exiting passes through a flow meter (2)
into the compliance chamber (3). The chamber has an elastic
membrane that distends under pressure, providing a dampen- Fig. 9 Control panel developed in LabVIEW to control the
ing effect on pressure waves. The resistor produces the back- pump parameters. The slider bars are used to select the desired
pressure through the occlusion of flow (4). A venous reservoir settings, and the “Update Settings” button actuates the com-
acts as a return tank for the loop and a supply for the pump (5). munication to the microcontroller to update the parameters.
A centrifugal pump has been included to study the effects of The gray box at the bottom shows the performance values
preload on the Harvard Pump (6). obtained from the microcontroller’s process monitoring.

045002-6 / Vol. 6, DECEMBER 2012 Transactions of the ASME

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 and hydraulic. Prototyping the design in this environment allowed ality of constant field voltage, with the ability to change the
for in silica evaluation prior to prototyping, and fine tuning the armature voltage for speed control. It also models the nonlinear
model with parameter estimation provided the high-fidelity com- speed-torque curve that is characteristic of shunt wound motors.
putational model for controller development. SolidWorks (Dassault Systemes, Paris) was employed in the
The physical model contains the subsystems pertaining to reverse engineering of the driveline (Fig. 2). The SimMecha-
the electrical voltage supply, motor model, driveline of the nicsTM Link allowed the SolidWorks geometry to be imported
pump, and the hydromechanical interface of the pump head. The into SimulinkTM as a SimMechanicsTM body, interfaced between
electrical voltage source was developed for this project using the mechanical elements of the motor driveshaft and the translat-
R
the SimElectronicsV toolbox and the Foundation Library ele- ing hydromechanical element representing the pump piston
ments. Particular design work was performed on the inrush current (Fig. 7). This model allows the change of the pivot point location
limiters (ICL) used to regulate the initial current supplied to the that dictates the stroke volume, creating an accurate working
smoothing capacitor. The desired performance was a fast charge representation of this driveline. The pump head model was con-
of the capacitor, while preventing an overload on the outlet pro- structed with elements from the SimHydraulics toolbox to repre-
viding the AC voltage. sent the geometries and fittings present. Included are sections of
The model of the shunt wound model delivered in the flow path size transition, ball valves, and sudden flow path size
SimscapeTM package does not allow for the independent variation changes. The values used in the model were derived from direct
of the armature and field voltages. The motor model had to be measurement of the pump head geometry to provide initial esti-
developed from literature and basic elements of the SimscapeTM mates, and parameter estimation was used to reach the final
Foundation Library (Fig. 6) [11]. The final model is presented in values.
Fig. 5, with the variables for the elements being determined The final model produced an accurate depiction of the updated
directly through experimentation or by means of parameter esti- Model 1423 operation. It allowed for sensors to be placed in the
mation. This shunt wound numerical model provides the function- computational model to better understand the operation of the

Fig. 10 Pump output with a staircase change in heart rate, while maintaining 50%
systole, and a stroke volume setting of 50 mL. The lower trace indicates the read-
backs from each of the parameters, which were used as inputs to the computa-
tional model. The spurious values in percent systole and BPM at some of the
transitions are due to the change to the new speed values midpumping cycle.

Journal of Medical Devices DECEMBER 2012, Vol. 6 / 045002-7

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 electrical and mechanical subsystems during operation to ensure The system fluid employed in this test was 40% glycerin in
their proper design. The input parameters for the model are power water at a temperature of 23  C. The viscosity of the fluid was
switch, stroke volume (mL) desired, heart rate (BPM), and percent verified at time of testing to be 3.707 cSt, and the density was con-
systole. It was desired that the inputs for the model would be anal- firmed to be 1.1047 g/mL. Both of these values are consistent with
ogous to the device settings for ease of testing and for future published physical property information on this solution [12–14].
determination of operating points for prescribed conditions. Compliance was maintained at 0.8 mL/mmHg throughout the
performance evaluation; the low value was chosen to keep the
2.5 Mock Circulatory Loop. The pump performance was pressure values safely under 150 mmHg throughout the various
validated in a mock circulatory loop (MCL) used to simulate conditions tested. The peripheral resistance was kept constant
systemic circulation conditions (Fig. 8). The MCL utilized is able throughout the trials at an effective flow path diameter of
to change left ventricular performance parameters, arterial com- 0.081 in2. These conditions sought to replicate the pumping loads
pliance, and peripheral resistance. The pump design discussed the pump would experience with the system fluid and isolate the
is responsible for the left ventricular simulation. An elastic pump performance by maintaining constant set points in the other
membrane-based compliance chamber is used to simulate the MCL components.
effect of arterial compliance in the system by means of distention The sensors utilized in the MCL were an M20XL (Transonic
in the membrane due to hydraulic pressure in the MCL. A hydrau- Systems Inc., Ithaca, NY) ultrasonic flow rate sensor downstream
lic proportional valve is utilized in the simulation of peripheral of the pump outlet and an EPX-V015P (Measurement Specialties,
resistance. The outflow of the peripheral resistor is returned to a Hampton, VA) pressure sensor monitoring the fluid pressure of
venous reservoir, which in turn supplies the hydraulic head and the compliance chamber. There was no interest in monitoring the
draw capacity for the pump. This closed loop MCL is effective in inlet pressure of the pump, as this portion of the system is not
simulating system circulation conditions and testing left ventricu- utilized in systemic circulation condition simulation. Previous
lar assist technologies. investigations have been conducted on the upstream pressure

Fig. 11 The 10% staircase change in percent systole from 30% to 70% and back
with constant heart rate and stroke volume. The spurious readbacks at the transi-
tions are due to the speed changes midstroke. These events are exacerbated in the
computational model due to the simulated system response to these perceived
rapid changes.

045002-8 / Vol. 6, DECEMBER 2012 Transactions of the ASME

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 effects of this pump, which further negated the need for pressure However, the pulse pressure remained consistent, due to the main-
sensors in this region [15]. tenance of the percent systole, constant stroke volume, and con-
stant arterial compliance value. This sensor data independently
3 Results and Discussions confirms the performance of the pump to that of the performance
data returned from the pump MCU.
3.1 Pump Motor Control Performance. The heart rate con- The speed set point maintenance in each phase of pumping
trol of the pump through the implementation of embedded control needed for accurate percent systole control was achieved, with
logic was successful, with a LabVIEW interface providing the set the performance illustrated in Fig. 11. The rapid adaption from
point control (Fig. 9). The ability of the MCU to control heart the feed-forward control utilized in the shaft speed control
rates over a range of values is shown in Fig. 10. The rapid change allows the rapid change in rpm needed for each phase transition,
to a new heart rate set point and the maintenance of that value is while ensuring the speed is maintained through that phase. This
clearly shown. Anomalies are present at set point changes, but are process control is shown to have a successful implementation for
consistent with the effects of changing speed midpumping cycle. this application and is capable of ensuring the performance of the
The slight variation in the readback values indicates that the sys- pump within the percent systole settings region of interest.
tem is capable of detecting the natural variability of performance The sensor data gathered from the MCL confirms the percent
in this mechanical system. systole change performance and consistency with this parameter’s
The flow rate and pressure responses of the MCL to only a sole effect on the system. The flow rate peaks increase in magni-
change in heart rate were consistent with this parameter’s effect. tude, but not frequency, which is consistent with the faster ejec-
The flow rate peaks increased in frequency and magnitude, which tion rate and maintained heart rate. The pulse pressure is reduced
validate the increased pumping frequency and rate of ejection. when nearing 70% systole; pressure development in the system is
The average hydraulic pressure rose proportionally to the change impacted due to the low systolic flow rate. This is consistent with
in the heart rate, as would be expected with a change in cardiac the constant compliance, unchanged stroke volume, and constant
output against a constant compliance and peripheral resistance. resistance. This confirmation of performance in MCL ensures that

Fig. 12 The functionality of the stroke volume change is illustrated by the continu-
ous change from 20 mL to 95 mL, with heart rate and percent systole remaining
unchanged

Journal of Medical Devices DECEMBER 2012, Vol. 6 / 045002-9

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

 the readbacks from the pump MCU are consistent with the diac function is important in developing higher functionality in
hydraulic sensor data. the designs of cardiac assist technology. An important facet of the
upgrade is the performance readbacks from the device during
3.2 Stroke Volume Control Performance. The stroke vol- operation. This information is more useful than previous methods
ume actuation and control was confirmed by the results in Fig. 12. of relating results to the set points of the device. Deviation from
This performance validation illustrates the ability to achieve the set points during operation are not a concern now, as the data
stroke volume values across the broad range provided by the obtained is relatable to precisely determined readbacks on the per-
mechanical design of this pump. The seamless return of stroke formance of the pump. The development of the LabVIEW inter-
volume values in this illustration indicates the resolution in posi- face integrates the local upgrades of the pump with the larger data
tion, even during movement of the actuator, which is able to be acquisition system for the mock circulatory loop into which this
achieved. The characterization of the stroke volume mechanism pump is incorporated.
was integral to this precise control. It was found that a second Limitations of the current design pertain to the speed of the
order polynomial fit (R2 > 0.99) provided a sufficient means of motor elements present in the pump. The shunt-wound technology
correlating the distance traveled from the 15 mL stroke volume employed for the pump motor is antiquated, and the motor is not
position to the stroke volume setting (Eq. (3)). as responsive as would be required for expanding the operating
region. The migration to a permanent magnet motor of similar
Stroke Volume ½mL ¼ 8:875x2 þ 14:192x þ 15:523 power characteristics provides the simplest upgrade in perform-
x ¼ distance of indicator in inches from 15 mL mark (3) ance. This motor technology is readily adaptable to the electrical
driving board currently employed. Actuation of the stroke volume
The change of stroke volume in the MCL was consistent with is not a rapid process, with beat-to-beat changes in set point not
easily realized. With an investigation of the maximum torque
this parameter’s effect in the system. The flow rate peaks
required for this application, a faster motor could be selected. The
increased in magnitude proportionally to the change in stroke vol-
ume. The average pressure and pulse pressure increased due to the refresh rate of the control elements would need to be increased to
account for this speed change; this is a minor code revision of the
change in cardiac output and ejection rate. The nonlinear change
in the conditions coincides with the shape of the stroke volume present architecture. The current stroke volume actuation involves
the disassembly of the device, which is not advantageous with the
curve, which in turn is effected by Eq. (3) and the constant speed
application of this upgrade to other pumps. A new technique of
of the stroke volume motor. The equivalent translational speed of
the stroke volume pivot point determined from the motor speed applying the stroke volume control should be investigated to allow
for a bolt on mechanism for this actuation.
and gear ratio was 3.21 in/min. The unimpeded movement of the
stroke volume movement during pumping validates that the stroke The success of this upgrade to the functionality of this pump
has been quickly realized in the expanded capability of data col-
volume can be changed in real-time and that the actuation method
lection in the mock circulatory loop. The rapid transitions to new
is not underpowered. Stroke volume actuation is integral to simu-
lating changes in cardiac output; the illustrated results from this conditions, and the capability of battery testing, have provided a
wealth of functionality in the investigations employing this
validation indicate this modified design’s capability in performing
those simulations. pump. Data resolution has increased with the elucidation of the
precise performance of the pump during these tests, and transi-
tions between settings has allowed for dynamic analysis of the
3.3 SimulinkTM Computational Model. The computational technologies being evaluated. Convenience of the LabVIEW
model developed in the design of the process control of this pump interface positions this upgrade as a user, and automation,
was found to have a high agreement with reference to the physical friendly design.
system. Utilizing the readback information from the pump
recorded from the in vitro experiments, these data traces provided
the input settings time course for the computational model. The References
model was able to accurately simulate the changes in heart rate
[1] Olin, C., 1971, “Pulsatile Flow Studies of Prosthetic Aortic Valves,” Scand.
(Fig. 10), percent systole (Fig. 11), and stroke volume (Fig. 12) Cardiovasc. J., 5(1), pp. 1–12.
conducted by the pump. The spurious readback values present at [2] Yeboa, K. C., Camp, T. A., and Figliola, R. S., 2009, “Modeling of Free Pulmo-
the settings changes, due to the change mid-heart beat, increased nary Regurgitation and Analysis of Mock Circulatory System,” 2009 Biomate-
the error in the computational model. This error was due to the rials Day, Clemson University, Clemson, SC, p. 25.
[3] Wolf, L., and Clinch, J. M., 1972, “Mock Circulatory System for Intra-Aortic
erroneous readbacks interpreted as sudden settings changes, which Balloon Testing,” IEEE Trans. Biomed. Eng., 19(1), pp. 38–46.
the computational model responded to. Another source of error in [4] Harvard Apparatus, 1981, Series 1400 Pulsatile Blood Pumps User’s Manual,
the computational model is the rounding present in the settings Harvard Apparatus Inc., Natick, MA.
communications from the pump. The settings are rounded to the [5] Harvard Apparatus, 2003, Model 665 Intermediate Ventilator User’s Manual,
Harvard Apparatus Inc., Natick, MA.
nearest integer; this dampens the deviations in values and causes [6] Eckhardt, H., 1998, Kinematic Design of Machines and Mechanisms, McGraw-
discontinuities. This issue is largely present in the percent systole Hill, New York.
and heart rate simulations; the stroke volume has enough resolu- [7] Microchip Technology, 2009, Compiled Tips ‘N Tricks Guide, Microchip
tion with the rounding for this to not have a noticeable effect. Technology Inc., Chandler, AZ.
[8] The MathWorks, 2010, “Simscape 3.0 User’s Guide,” The MathWorks, Natick,
Although the model does not have perfect agreement with the MA.
experimental data, the performance is such that it is a capable tool [9] Microchip Technology, 2009, PIC18F2455/2550/4455/4550 Data Sheet,
in pre-evaluating settings and able to be incorporated into compu- Microchip Technology Inc., Chandler, AZ.
tational models of the mock circulatory loop. [10] Microchip Technology, 2005, MPLAB(R) C18 C Compiler Getting Started,
Microchip Technology Inc., Chandler, AZ.
[11] Toliyat, H. A., and Kliman, G. B., 2004, Handbook of Electric Motors, CRC,
4 Conclusion Boca Raton.
[12] Segur, J. B., and Oberstar, H. E., 1951, “Viscosity of Glycerol and Its Aqueous
The ability to remotely actuate condition changes, read back Solutions,” Ind. Eng. Chem., 43(9), pp. 2117–2120.
performance values during operation, and use of embedded logic [13] “Density of Glycerine-Water Solutions,” Dow Chemical Corporation, 2011,
to maintain set points creates a device capable of integration into http://www.dow.com/glycerine/resources/table4_3140.htm
[14] Dow Chemical Corporation, 2011, “Viscosity of Aqueous Glycerine Solutions,”
fully automated testing systems. Rapid updates to the set points of http://www.dow.com/glycerine/resources/table18.htm
operation on the pump expand the investigative possibilities, [15] Dorsi, J., 1990, “Testing a Multiple Disk Pump for Use as an Artificial Ven-
including dynamic analysis. This evaluation of transitions in car- tricle,” MS thesis, Texas A&M University, College Station, TX.

045002-10 / Vol. 6, DECEMBER 2012 Transactions of the ASME

Downloaded From: http://medicaldevices.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use