Multidisciplinary Senior Design Conference
Kate Gleason College of Engineering
Rochester Institute of Technology
Rochester, New York 14623
Project Number: P15318
GASEOUS MASS FLOW RATE CONTROLLER
Brian Church
Rochester Institute of Technology
Mechanical Engineer (Co-Team Lead)
Anthony Salmin
Rochester Institute of Technology
Electrical Engineer (Co-Team Lead)
Tyler Breitung
Rochester Institute of Technology
Mechanical Engineer
Ross Bluth
Rochester Institute of Technology
Electrical Engineer
Michael Oplinger
Rochester Institute of Technology
Mechanical Engineer
Stephen Mroz
Rochester Institute of Technology
Electrical Engineer
ABSTRACT
A Gaseous Mass Flow Rate Controller (GMFRC) is a mass flow regulating device capable of metering
precise amounts of compressed natural gas (CNG). In the early 2000s, prototypes of such a device were first
developed and tested by Dr. Roman Press. Although the prototypes demonstrated high accuracy and fast response
times, the main functionality that they both lacked was that they were not stand-alone devices and had to be operated
with external equipment. The main objectives of this project are to make a fully functional, stand-alone prototype
that further increases the overall accuracy and performance, as well as develop a means to mix the regulated fuel
with air and deliver it for use in an internal combustion engine.
INTRODUCTION
As oil and gasoline prices continue to rise, along with their environmental concerns, the discovery of
natural gas sources in the United States creates the opportunity to utilize natural gas as an alternative for
transportation fuel. In order to harness the energy in CNG, a system that can precisely regulate and deliver the fuel
to an internal combustion engine is needed.
Currently, there are a few manufacturers of systems that utilize CNG for use in an internal combustion
engine. Bosch’s Bifuel CNG-System was first introduced in 2004, when it became available with the Chevrolet
Astra in Brazil [1]. More recently, Ford’s Bi-Fuel CNG/LPG Engine Package became available on their F-Series
line of trucks in 2014 [2]. The advantages of both systems are that they can operate on either CNG or gasoline alone
and do not suffer noticeable differences in torque or horsepower between the two fuels. The major disadvantage of
the systems is that they are complicated and expensive. Ford’s Bi-Fuel CNG/LPG Engine Package, the more
Copyright © 2015 Rochester Institute of Technology
expensive of the two, costs between $6000-9500 on top of the cost of the truck. The reason for the elevated cost is
that these systems utilize separate fuel injectors and fuel lines for each of the fuels, which leads to increased costs
due to added components and more machining time that needs to go into the engine block. This also means that
these systems cannot be retrofitted to an existing engine platform without a large number of modifications.
With the intention to make a simpler and less expensive fuel regulating system for CNG, the original
GMFRC prototype was designed and developed by Dr. Roman Press in 2001. The main objective of the prototype
was to show proof of concept of a device that could regulate fuel just as well as the more complicated and expensive
competition. This prototype was very successful, as it was able to regulate the fuel with an accuracy of ± 1 % of full
scale and with a response time of less than 70 ms [3]. However, this design also had a few deficiencies; mainly the
fact that it was not ready to be used with an internal combustion engine. It did not have its own dedicated controller
and did not have any means of mixing the regulated fuel with air or delivering it for use in an internal combustion
engine. Also, the device was not designed with the intention to be mass produced.
The main objectives of this project are to make a fully functional, stand-alone GMFRC prototype that
further increases the overall accuracy and performance, as well as develop a means to mix the regulated fuel with air
and deliver it for use in an internal combustion engine.
PROCESS
The team was tasked with developing the next generation of the GMFRC over the planning and design
stages of Multidisciplinary Senior Design. The planning stage involved using input from the customer, Dr. Roman
Press, to draft a list of customer requirements and quantifiable engineering requirements. Once the scope of the
project was understood through the requirements from the customer, concept generation started. Several concepts
were developed for subsystems using various different tools learned throughout the early portions of the semester.
Once concepts were generated, a design review was held with our team guide. From the design review, optimal
concepts for the subsystem were chosen. Once the subsystems were developed, the integration of subsystems into a
complete system started with detailed design. Several weeks of engineering discussions and analysis allowed for
complete system integration. After the team had a final design review with our guide and customer, we obtained
approval to move onto the build phase.
rqmt. # Importance
Customer
Rqmt. #
CR1
CR2
CR3
CR4
CR5
CR6
CR7
CR8
CR9
CR10
CR11
CR12
Importance
9
9
9
9
9
3
3
3
3
3
1
1
Description
Accurate and repeatable mass flow rate control
Operate on the nominal OEM control board power
Usable with gaseous fuels
Wide dynamic range of flow control
Low production cost
Integrated package design
Long life in an automotive environment
Stable to shock and vibration
Small and lightweight
Fast opening/closing response time of actuator
Distribution plate efficiently mixes fuel with air
Usable with a midsize engine
TABLE 1: CUSTOMER REQUIREMENTS
ER1
ER2
ER3
ER4
ER5
ER6
ER7
ER8
ER9
ER10
ER11
ER12
ER13
ER14
9
9
9
9
9
9
9
3
3
3
3
3
1
1
Source
CR1
CR1
CR1
CR2
CR3
CR4
CR5
CR6
CR7
CR9
CR9
CR10
CR11
CR12
Engr. Requirement
Accuracy of mass flow rate control
Repeatability of mass flow rate control
Leak rate of device
Operating voltage
Usability with gaseous fuels
Dynamic range of flow control
Cost of production device
Number of separate assemblies
Temperature limit
Size envelope of production unit
Total weight of production unit
Open/close response time
Fuel mixing shape
Compatible engine size
Unit of Marginal Target
Measure Value
Value
%
±5
±1
%
±5
±1
cc/min
25
25
Volts
8-16
6-18
Yes/No
Yes
Yes
Ratio
50:1
100:1
$
100
100
#
3
2
°C
75
80
in3
96.5
86.5
lbs
5
4
ms
50
50
Target No Vortex Vortex
Liters
4.7
4.7
Ideal
Value
± 0.5
±0
0
5-24
Yes
>100:1
<100
1
85
72
3
<50
Vortex
All
Will be acknowledged in the design, but not testable:
ER15
3
CR7
Operating life
ER16
3
CR7
Number of full cycles
ER17
3
CR8
Shock resistance
ER18
3
CR8
Vibration resistance
TABLE 2: ENGINEERING REQUIREMENTS
There are three major components that contribute to the design of a successful GMFRC device. The first is
the design of the fuel metering mechanism to allow gaseous flow through the device. The second is the design of the
controller and supportive electronics to accurately control the flow metering mechanism. The last is the fuel delivery
system that will be used to mix the fuel with air and deliver it for use in an internal combustion engine.
MECHANICAL DESIGN
To satisfy the first component of this project, thermodynamics and gas dynamics must be understood in
order to develop a method to regulate mass flow. From a thermodynamic aspect, it is known that pressure and
temperature have a direct relationship with each other if there is a control volume, meaning that volume is constant
in the situation being analyzed. With the knowledge of how pressure and temperature act together, it is imperative
to understand how they will affect mass flow rate, which brings in gas dynamics to explain mass flow rate.
Proceedings of the Multi-Disciplinary Senior Design Conference
Mass flow rate at a high level is the product of density, velocity, and area. When breaking these
components up, it is recognized that velocity is proportional to the square root of temperature and density is
proportional to pressure. The equation for mass flow rate can be derived by combining the equations of velocity and
density when in terms of pressure and temperature, as shown in Equation (1).
̇
√
(
)
(
)
( )
One last thing to consider when designing the valve to control mass flow is the possibility that the flow will
become choked. Choked flow occurs when any gas property exceeds the critical value. These properties include the
density and pressure of the flow as well as the area of the opening that the flow will be going through. Once a flow
is choked, the resultant mass flow is equal to the product of density, area, and the speed of sound ( =1). In this
project, choked flow can be achieved through the variation of pressure. Now that there is an understanding of how
the gas will behave, it is critical to design a valve that will allow for accurate and repeatable mass flow rate control.
The valve on the GMFRC needed to be designed so that it could provide a tight seal in the fully closed
position in order to reduce the leak rate of unwanted fuel being delivered to the engine. To accomplish this, an
aluminum rotating disk along with a delrin output port were used. Figures 1a and 1b show the valve in the fully open
and fully closed positions, respectively. When the slot in the disk overlaps with the hole through the output port,
flow is allowed through the output port. When there is no overlap, the flow is restricted from entering through the
output port. The mating surfaces between the rotating disk and output port both had spherical profiles defined by the
same radius. To achieve a sealing configuration, the rotating disk used an outside radius and the output port used an
inside radius so that they would match up when pressed together. Aluminum was chosen for the disk since it had a
rather complex profile that had to be Computer Numerical Control (CNC) machined. The material of the output port
needed to be softer than aluminum to promote the wear of the output port instead of the disk. It also had to be a
material with a relatively low coefficient of friction against aluminum so that the valve would have as little friction
as possible. For that reason, delrin was the material of choice. To keep the output port pushed up against the rotating
disk, a standard compression spring was used. The benefit of this is that the valve will still have a tight seal, even as
the delrin wears over time.
FIGURE 1a: VALVE IN OPEN POSITION
FIGURE 1b: VALVE IN CLOSED POSITION
The distribution plate needed to be designed with a few key requirements in mind. It had to be able to
evenly mix the regulated CNG with air coming into a combustion engine through the air intake. This device should
be located between the throttle body and the air intake, mounted in a way that allows it to mix the fuel with the air
passing through the throttle body. To limit the amount redesigning that would be needed to retrofit an existing
engine, the overall thickness of the device had to be smaller than 0.5 in. The design we used is a ring-shaped plate
that directs the working fluid through eight small nozzles. The nozzles are angled radially and tangentially in order
to induce a “swirling” effect, as shown in Figure 2a. This effect was first evaluated using SolidWorks’ built in
Computational Fluid Dynamics (CFD) simulation toolbox. The results, shown in Figure 2b, were very promising
and provided a significant “swirling” effect.
Copyright © 2015 Rochester Institute of Technology
FIGURE 2a: DISTRIBUTION PLATE
FIGURE 2b: CFD ANALYSIS
ELECTRICAL DESIGN
On the original GMFRC, Dr. Press used the CTS 640 series Rotary Actuator to open and close the valve
used to regulate mass flow rate of the fuel. The actuator is a 4-pole torque motor with a return spring feature, a
milled male “D” shaped keyway, and can be operated with 0-5 A. The actuator is commonly used in automotive
transmissions, which makes it manufacturable in large quantities, inexpensive, and able to operate in the -40-85° C
temperature range. The torsion return spring acts as a safety feature to close the actuator in the event of power loss
and cut the flow of fuel. Based on the preceding features, it was decided to incorporate the CTS Rotary Actuator into
the new design.
To control the flow of the GMFRC, it was decided that a controller with proportional feedback would be
implemented. The controller needed to have interface circuitry to read a pressure sensor, a temperature sensor, and a
position sensor for feedback control as well as supporting circuitry to power these sensors, filter the sensor outputs
and control the actuator. The circuit architecture, shown in Figure 3, is basic layout for the circuit hardware and
signal paths.
FIGURE 3: CIRCUIT ARCHITECTURE
Proceedings of the Multi-Disciplinary Senior Design Conference
Proportional feedback was selected since it is a simple to implement and is not computationally intensive,
which would help preserve the response time of the actuator. To implement proportional feedback, the Teensy 3.1
development board was selected for a few reasons. The first being that a microcontroller was needed to make the
GMFRC a stand-alone device. The second being that the development board includes interface circuitry, making it
easier to use and design supporting electronics. The third reason is that the Teensy 3.1 has a clock speed of 72 MHz,
256 KB of flash memory, USB interface, and 8 Timer interrupts. These features made the Teensy 3.1 much simpler
to interface with and program compared to the other development boards and stand-alone microcontrollers that were
researched. The fourth reason is that the Teensy 3.1 could operate in the desired -40-85° C temperature range. The
last reason is that it is relatively inexpensive for a development board.
For the rotary position feedback, the Delphi 514 Series rotary position sensor was selected. The position
sensor is an analog output, 5 kΩ ± 40% potentiometer that has a female receptacle that accepts the “D” shaped
keyway in the shaft of the rotary actuator. This sensor is commonly used in throttle control for vehicles, making it
manufacturable in large quantities, inexpensive, and able to operate in the -40-85° C temperature range requirement.
The position sensor can take a wide range of input voltage with an output of 0.925% of Vin per degree, but is
recommended to operate at 5V or higher. To ensure proper interfacing with the Teensy 3.1, the position sensor was
powered with a 5V supply from an On Semiconductor LM317MBSTT3G adjustable voltage regulator and the
output was run to a one half voltage divider. This is to guarantee the sensor output is 2.5V or less and ensures that
this will not damage the microcontroller during operation.
For the pressure sensor feedback, the Measurement Specialties M7139 Pressure Transducer was selected
for its suitability to work in liquid and gaseous media. This sensor is commonly used in vehicle and industrial
applications, making it manufacturable in large quantities, inexpensive, and able to tolerate the -40-85° C
temperature range requirement. The M7139 is an analog output, gauge pressure transducer capable of operating up
to 200 PSI (13.79 Bar) with an accuracy of ± 0.25% of full-scale. This operating range well exceeds our requirement
of 90 PSI (6 Bar). The M7139 is powered with a 5V input from the LM317MBSTT3G and its output ranges from 05V. Therefore, the output of this sensor was run through a one half voltage divider to guarantee a sensor output of
2.5V or less to ensure no damage is done to the Teensy 3.1.
For the temperature sensor feedback, the Delphi 150 Series Coolant Temperature Sensor was selected for
its easy to mount package design. Since this sensor is specifically designed for automotive applications, it is
manufacturable in large quantities, inexpensive, and able to operate in the -40-85° C temperature range requirement.
The coolant temperature sensor has an accuracy of ± 0.6° C in the desired temperature range. It uses a thermistor
pulled up to 5V power through a resistor to the LM317MBSTT3G and the output ranges from an analog 0-5V. The
output of this sensor was ran through a one half voltage divider to guarantee a sensor output or 2.5V or less to ensure
no damage is done to the Teensy 3.1.
To ensure the Teensy 3.1 properly reads output signals of the feedback sensors, the position, pressure, and
temperature sensors were run through a unity gain operational amplifier (op-amp) to isolate the output resistance of
the voltage divider and the input resistance of the Teensy 3.1. Since these resistances are on the same order of
magnitude, directly connecting the feedback sensor voltage dividers to the Teensy 3.1 would cause a loading effect
on the output signals and reduce the signal range. The unity gain op-amp does not amplify the signal output, but
provides a high output resistance, which renders the loading effect on the sensors as negligible. A MCP6004 quad
operational amplifier was used to make the unity gain op-amps for the feedback sensors and input voltage sampler.
The MCP6004 was selected for its rail to rail operation, its operating temperature range of -40-85°C, and its low
cost.
The Teensy 3.1 uses the feedback sensors above to control the actuator position of the rotary actuator by
pulse width modulating (PWM) the input voltage to the actuator. The larger the “on time” of the PWM signal, the
more current the actuator draws and the more the shaft rotates. The shorter the “on time” of the PWM signal, the less
current the actuator draws and the less the shaft rotates. To meet the current demand of the actuator, an H-bridge
was used as a digital relay to PWM the actuator input voltage, while also isolating the Teensy 3.1 from the back
EMF generated from operating the actuator. The Freescale MC33931EK was the selected H-bridge for a few
reasons. The first is that the MC33931EK can handle up to 5A, which exceeds the maximum required current
needed to operate the actuator. The second is that it continuously operates from 0-40V, which easily includes the
input voltage range specified in the customer requirements. The third reason is that it can operate in the desired
temperature range of -40-85°C. The last reason is that it can be operated using 3.3V, which is the logic output of the
Copyright © 2015 Rochester Institute of Technology
Teensy 3.1. Figure 4, below, summarizes the electrical design for both the circuit architecture and electronic
components being used.
FIGURE 4: CIRCUIT SCHEMATIC
FINAL SYSTEM PROTOTYPE
Combining all of the subsystems together yielded the final system prototype shown in Figure 5. The
pressure and temperature sensors are connected to the input of the device on the left, while the output port subassembly is connected on the right. The rotary actuator’s shaft is connected to the rotating disk through the top,
while the position sensor mates with the end of the shaft from the bottom. To consolidate space and provide
adequate heat dissipation for the H-bridge, the electronics were placed on a double sided printed circuit board
(PCB), as shown in Figure 6.
FIGURE 5: EXPLODED VIEW
Proceedings of the Multi-Disciplinary Senior Design Conference
FIGURE 6: PRINTED CIRCUIT BOARD
RESULTS AND DISCUSSION
The major engineering requirements that we were unable to meet were the leak rate of the device in the
closed position, the cost of a production device and the response time of the device, as shown in Table 3. Although
we could meet the leak rate requirement of less than 25cc/min for pressures below 10 psi, the leak rate increased
significantly at pressure above this. The reason for this was because a compromise between leak rate and being able
to turn the actuator was needed. A strong spring that kept the leak rate very low caused too much opposing torque
for the actuator to overcome. A weak spring allowed the actuator to easily move, but caused a very high leak rate at
any pressure.
Next, we were unable keep the estimated cost of the production GMFRC below the $100 metric. However,
this is a conservative estimate since many components in the bill of materials we interpolated to yield a reasonable
estimate. It is possible to reduce the price of the production GMFRC by obtaining proper quotes on prices, which
was out of the scope of this project.
Lastly, we were unable to meet the response time requirement of 50 ms or less. Since the actuator response
time of fully opening or closing the actuator was less the 10ms, the problem lies within how the control algorithm is
being implemented in the code on the microcontroller. The response time can be slowed by an additional 1-2
seconds depending on the amount of friction between the output port and rotating disk and by the magnitude of the
normal force that keeps the output port pushed against the rotating disk.
There were also a few engineering requirements in particular that we were able to meet the marginal value
for but not the target value. We were not able to directly test the accuracy and repeatability of the device with
regards to flow rate control, as a measurement device more accurate than our specification was not able to be
obtained. Therefore, we evaluated these metrics based on the device’s ability to be controlled to a specified angular
position since that directly correlates to the flow rate. Since the control algorithm decides when the angular position
is settled by seeing if it is within a specified percentage of the commanded position, it was expected that the
accuracy would be constant and the repeatability would vary. Contrary to these expectations, our test data showed
that the repeatability was very high and the accuracy was only marginally acceptable. Most of this error attributed
to the accuracy was accumulated in the higher ranges of operation, being 50° or higher. Since the observed positions
were much further from the commanded positions than in the lower range, this problem could be caused by
the position sensor not being properly calibrated at the higher position ranges, and thus led to the increase in the
error for this test.
Lastly, the dynamic range that we obtained from the device was just over 75:1. This problem came from
the fact that the slot in the disk was made by a grinding process instead of Electrical discharge machining (EDM), as
original planned. The grinding process was not able to produce a sharp point at the end of the slot as an EDM
process would. This resulted in a higher minimum flow rate and therefore a decreased dynamic range.
Copyright © 2015 Rochester Institute of Technology
TABLE 3: PERFORMANCE VS ENGINEERING REQUIREMENTS
CONCLUSIONS AND RECOMMENDATIONS
A closed-loop feedback control was successfully developed and implemented, allowing the GMFRC to
operate as a standalone device. The system offered control with a reasonable accuracy and a high precision, however
its response time was much slower than required. From a mechanical standpoint, the tolerances and surface finishes
on the components were not sufficient to reduce the leak rate and sliding friction to acceptable values. The
distribution plate, on the other hand, successfully demonstrated proof of concept as a method of introducing and
mixing CNG into the intake air of an automobile.
Looking to the future, there are several things that can be done to improve the current design, both
electrically and mechanically. On the electrical side; the code implementation can be modified to improve timing
and accuracy. The idea of using a proportional-integral (PI) controller instead of the current proportional (P)
controller could be pursued. The pressure and temperature sensors also need to be integrated into the control
algorithm. The PCB can be redesigned to reduce size and cost, and improve overall capability of the device.
Looking at the mechanical design; a new housing should be made that incorporates the sensors and PCB, reducing
the overall size of the device. Given the shortcomings of the current design, a valve design that uses a cam and ball
may be worth pursuing. Further development of the distribution plate can include making it suitable for mass
production and implementing it in tests involving a throttle body and air manifold from an internal combustion
engine. Lastly, a new calibration of the device would be needed to make sure the device works properly with
varying input pressure and temperature.
REFERENCES
[1] Bosch Mobility Solutions, n.d., “Bifuel CNG-Systems.” From http://products.bosch-mobilitysolutions.com/en/de/powertrain/powertrain_systems_for_passenger_cars_1/bifuel_cng_saugrohreinspritzung_1/
bifuel_cng_systeme.html#
[2] Ford Motor Company, 2013, “First CNG-Capable 2014 Ford F-150 Rolls Off The Line in Kansas City.” From
https://media.ford.com/content/fordmedia/fna/us/en/news/2013/11/21/first-cng-capable-2014-ford-f-150-rollsoff-the-line-in-kansas-c.html
[3] Press, R.J., 2001, “Quantum Fuel Metering Valve Controller Development,”n.p., n.p.
ACKNOWLEDGEMENTS
We would like to thank our guide, Edward Hanzlik, for all of his help and support throughout the entire design
process. We would also like to thank our customer, Dr. Roman Press, for his assistance with our system design, Dr.
Lynn Fuller for his manufacturing services and assistance with our electrical hardware, and both the Mechanical
Engineering Machine Shop and Brinkman Lab staff for their assistance with machining our components.