Final Report

Balanced^3

Group 15

Columbia University

Department of Mechanical Engineering

Senior Design

May 3, 2017

Maciej Biernacki, Will Church, Jonathan Cohen, Tom Rasmussen

 Table of Contents

Executive Summary 2

Introduction 3

Background Information 4

Design Overview 5
Mechanical Layout Design 5
Parameter Identification 7
Material Selection & Manufacturing 7

Analysis 11

Design Details 12
Cost 13

Experiments & Test Results 14

Conclusions 15

Appendices 16

1
Executive Summary

A stand-alone section that describes the product, its intended market and the manner in which it should be
used. This section must describe in detail those factors that had the most significant influence on the final
design. The discussion must highlight the strengths and weaknesses of the concept and indicated all critical
technical, economic or aesthetic issues.

● Should be a synopsis of the ENTIRE REPORT.

○ Brief Introduction.
○ Background, Design, Conclusions.
○ Exactly like the Abstract of a paper.
● Do not treat the Executive Summary as an introduction to the work (even though it is the first section
of the report).
● In industry, this is the most important section since it is the one that everyone in the organization will
read.
● It is usually easiest to write this section last, after the rest of the report is written.
● Do not use equations or symbols or figures.
○ Describe everything in words.
● No details.
● At most 1 page.

2
Introduction

This project is primarily concerned with the field of control theory as it applies to

complex dynamic systems. When viewed as a demonstration in control theory, it has a wide

variety of potential applications and related systems. Digital control is becoming increasingly

important and has importance in industrial applications requiring precise, automated control of

systems. These systems include manufacturing equipment, mechatronic systems, and aerospace

systems. Knowing how to design dynamic systems implementing digital control for optimal

performance is a critical skill for the modern mechanical engineer.

The balancing cube is an important demonstration in control of dynamic systems because

it provides a very tangible and observable platform for learning how design parameters lead to

the performance of a system that takes delicate tuning and balance. In the design, we were

primarily tasked with analyzing tradeoffs of several design parameters including frame weight vs

rigidity, wheel weight vs inertia, motor weight vs torque, motor speed vs torque and sensor

accuracy vs frequency. In tuning the performance of the balancing cube, we were tasked with

derivation of equations of motion, system identification, and optimal pole placement. These

procedures in mechanical, electrical, and control systems design are common for any complex

system, and we are confident that the experience we have gained from these designs will be

applicable to the real world applications stated above.

The main goal of the project was to create a cube that was able to converge quickly to a

steady, balanced state with zero wheel velocity even in the presence of significant external

disturbances, and to remain stable in this configuration indefinitely in the absence of these

3
disturbances. The performance benchmark was to have a recovery angle of ±10° from

equilibrium and to minimize oscillatory behavior due to mechanical or control system problems.

The design that follows, as well as the demonstration of the cube, show that we have achieved

these goals.

Background Information

The literature review for this project uncovered various resources that motivated our

development of the cube. The most relevant information came from several papers from ETH

Zurich that described the development of the Cubli, a similar self balancing cube device. These

papers were an inspiration for the concept of our project. We expected their details to be more

helpful, but did not end up implementing any of the same control algorithms that they used. The

primary use of this project was for general concept generation.

The patent search portion of our literature review revealed several applications in

industry which used similar control systems. In particular, a patent for ​“Powered unicycle”

(​US20070158117A1), an electric unicycle which demonstrates ​a “control system arranged to

automatically maintain the fore-aft balance of the unicycle via operation of the motor,” was

applicable to our project. In general, we found in our search that satellite attitude control systems

often use reaction wheels or gyros to control their orientation, and we found studying these

concepts helpful in understanding the operation of our project. More terrestrial examples that

helped our understanding were generally related to inverted pendulum problems or Segway-like

balancing vehicles. Since the control for any system is unique to its dynamics, we found the

textbook “Linear State-Space Control Systems” to be useful in deriving and tuning our control

4
law.

Design Overview

Mechanical Design

Preliminary mechanical design focused on methods for joining the six faces of a cube in a

way that fully constrains the faces and provides sharp corners and edges on the exterior of the

cube. For previous designs which were considered, see Appendix A. The final design came from

the realization that each of the six faces of the cube only needs to provide two edges each to sum

to the 12 edges of a cube. This resulted in the face shape seen in the figure below.

Figure X. Frame Design

The designated corner pieces allowed for consistent and secure mating of the faces of the

cube while still providing sharp points at the corners. Originally, these corner parts were

designed such that the screws converged in the center of the cube. This required a very small

5
length of threading in the corner part to allow fastening but prevent interference. As it became

apparent that these parts would be especially suited to 3D printing, this small threading became

unacceptable. The bolt pattern of each face was redesigned with an offset transverse to its edge

that allowed the bolts to overlap and be fastened with a nut on the other side of the cube. This

design required two different orientations for the corner piece, but resulted in a much more

secure joint.

Figure X. Corner Overlap

The design of the internal mechanisms followed, beginning with the overall placement of

components within the cube. The figure below shows rationale for the final design. Initially

considered was an intuitive configuration in which the motor mounted directly to the frame of

the cube. However, it was decided that the wheel would required a support at its other end to

prevent a cantilevered configuration that would be unstable. Having the wheel as close as

possible to the face also allowed its diameter to be larger without interfering with wheels on

adjacent faces.

6
 Figure X. Interior Design

Design of the reaction wheel began with the question of how to couple the shaft of the

motor to the axis of the wheel in a secure fashion. Initial designs considered a wheel made out of

a single piece of stock. This would have required much more complex manufacturing processes.

It was decided to create the body of the wheel from a single plate stock and affix connecting

pieces that couple the wheel to the shaft with set screws on one end, and mate the wheel to a

7
support bearing on the other. To ensure a tight, concentric fit between these parts, special

shoulder screws were used that came precision manufactured to be slightly undersize the

diameters of the holes in which they are inserted. To optimize the flywheel, as much material as

possible was concentrated at the exterior of the wheel, and as much material as possible was

removed from interior of the wheel. This is because the wheel needs to have a maximized

moment of inertia to prevent saturation of wheel velocity per a commanded torque, and a

minimized mass to keep the cube body inertia small enough to be easily affected by motor

torque. These considerations resulted in the highly optimized geometries of the connecting

pieces and the large pockets at in the interior of the main wheel body.

Figure X. Reaction Wheel Assembly Design

Design for assembly was carefully considered in the overall design of the cube. Because

smooth rotation of the wheel is required for reliable operation of the cube, it was ensured that a

8
motored-side subassembly could be entirely removed and replaced without detaching any

fasteners that would affect the mounting of the reaction wheel. In other words, each motor-side is

a complete subassembly that can be removed from the cube body easily for tweaking. This

proved to be a useful design when testing and assembling the cube.

Figure X. Motor Side Subassembly

Parameter Identification

Recall the equation for angular momentum (L) and torque (𝜏):

L = Iω

dL
τ= dt

9
Where I and ω are the moment of inertia and angular velocity about the axis of rotation,

respectively, and t is time.

Note that the conservation of angular momentum for rigid bodies requires that a torque applied

by the motor translates directly to a torque in the opposite direction on the frame. We can

therefore define a value of the maximum angular displacement of the frame from vertical for

which the frame can return to the vertical position by applying the maximum motor torque,

assuming zero initial conditions. This ​recovery angle​ θ​rec​ is derived from the equilibrium

between the torque from the motor and the torque supplied by gravity:

τ motor = mgl sin(θrec )

τ motor
θrec = sin−1 ( mgl
)

Where m is the total mass of the frame assembly, l is the distance from the axis of rotation to the

center of mass of the assembly, and g is acceleration due to gravity. This was an important

performance metric for specifying the motor.

Note that the recovery angle is independent of the moment of inertia of the wheel. This is

deceiving, because a higher moment of inertia of the wheel will result in a smaller increase in

angular velocity for a given torque. This is desirable because the torque capabilities of the motor

will diminish as its speed increases.

10
Using the conservation of energy, we can find the final ω for the reaction wheel for the recovery

angle scenario described above. The change in potential energy when tilted an angle θ​rec​ from

vertical is:

P = mgl(1 − cos(θ))

The kinetic energy of the wheel is:

K = 12 Iω 2

Equating these terms yields the final angular velocity of the wheel:

√
2mgl(1−cos(θ))
ω= I

Comparing this velocity to the maximum velocity of the motor was a useful design metric.

Material Selection & Manufacturing

Figure X. Machined Parts

The primary material considered for structural components of the cube was aluminum

due to its high Young’s modulus, relatively low density, and relatively high yield stress. Initially,

11
the alloy 7075-T6 was considered for its superior strength-to-weight ratio over other common

alloys. However, the cost of this alloy was prohibitively high compared to the more common

6061-T6, and the loading experienced by the structural components was not expected to

approach the yield stress of this material, even for a very small cross section. The frame needs to

primarily provide rigidity, and alloying has almost no effect on the stiffness of a material.

Figure X. Machining Wheel

Due to its high manufacturability rating, aluminum was initially specified for the reaction

wheels, but its relatively low density made the wheel cross section prohibitively large for fitting

into the envelope of the frame for the required inertia of the wheel. Therefore, low-carbon steel

(1018) was used. This material was dense enough to allow the wheels to be about 4 inches in

diameter and ¼” thick, which fit well in the cube.

12
 A significant effort was put toward design for manufacturing in the design of all

machined parts. The design contained 24 complex machined parts which necessitated a

methodical approach to manufacturing that made use of reusable fixturing and dimensions of

parts which are available to purchase as stock. For example, the sides of the cube were specified

to be ⅛” thick and slightly under 6” x 6” to allow quick machining from a standard stock size.

This also decreased stock costs. A jig plate was made (see below) to allow the complex contours

of the sides to be quickly machined in one program, and to prevent rezeroing of the machine for

each side.

Figure X. Fixturing for Side

The motor mounting plates, as well as the interface part between the side and the motor

plates, were designed such that multiple could be made from a piece of stock in a single

operation (see below). This decreased machining time dramatically by eliminating tool changes,

13
machine rezeroing, and re-fixturing. These considerations were critical to timely completion of

the manufacturing of the parts and allowed for more time dedicated to control tuning.

Figure X. Multiple Parts from Plate Stock

14
Analysis

● Summary of analyses performed.

○ Describe how the analysis was done.
■ List ALL the equations used (write out equations).
■ List any assumptions used.
■ List the values of any physical constants (give units).
○ All parameters and symbols in the equations should be defined.
○ You do not have to derive equations from first principles.
● Summary of modeling performed.
○ Describe the models.
■ What software was used?
■ List any assumptions made.
○ Show model results.
● Calculation of expected output for the system.

15
Design Details

● Include a schematic of the design.

○ Show all sensors and actuators (how is performance measured and how is the system
controlled).
○ Show important aspects.
○ This is different from the CAD drawings.
● You should not simply list the elements which appear on the schematic in the report body (the reader
can see what is on the schematic). However, you may want to expand on a specific aspect of the
system or a specific element with an additional sentence.
● Describe the different sections of the design (if appropriate).
● Describe the operation of the system.
● Describe any support equipment.
● Provide enough information on what you specifically did that someone else, not familiar with this
work, can understand your design.
● Include a cost of the work. You were limited in what you could purchase from department funds.
This is what you actually spent (see below).

The cube incorporated all aspects of Mechatronics by combining Mechanical Engineering,

Controls Engineering, Electrical Engineering and Computer Science. In this system our physical
structure and dynamics were physically models using mechanical engineering skills.
The following schematic, in Figure X, is of the built design that uses only one motor. The cube’s
rotation has been constrained to one axis since it is on an edge, it has only one degree of freedom
to rotate about. Although, this was named the 2D case since it incorporated the full cube
structure which was an advancement on a simple inverted pendulum.

(Share control of system)

16
 Figure X. Edge balancing electronics schematic (1D & 2D Cases)
The following schematic in figure x was developed for the 3D or corner balancing case. It has
the same electrical architecture as the 1D & 2D case except for there being three motor
controllers and brushless motors.

17
 Figure X. Corner balancing electronics schematic (3D Case)

Explain components
The STM32 microcomputer was used as the controller for the entire project.
Operation procedure

18
 Figure X. Control Logic (UPDATE CONTROL LOGIC JONNY)

Support equipment
The STM32 board was used as an independent microcontroller that did not need . The only
peripheral electronics that the cube needed was a voltage source that could provide between
12-24 Volts and a peak current of 6 Amps. We decided to use

19
Cost

The actual total cost of the cube is just under $900. The bulk of this cost (72%) is due to

the high cost of the brushless DC motors and their controllers. However, the performance of our

design is also largely attributable to the high torque-to-weight ratio of the motors and the

impressive performance of the motor controllers. The goals of the project would not have been

achievable without spending the bulk of our budget on these parts.

20
21
Experiments & Test Results

● Describe how the data was acquired. What exactly did you measure and how did you do it?
● How does this data demonstrate the operation of your design?
● Describe how the data was reduced.
● Any details should be put in an Appendix.
● Summarize the results (tables, plots, figures, descriptions if appropriate). The format for plots
generated by EXCEL is shown below.
● Discuss sources of experimental or analytical error.
○ Do not list human error. (This is implicitly understood – everyone makes mistakes.)
○ Restate assumptions and discuss if these are valid.
○ List possible measurement or equipment errors.
● Be quantitative and definitive (have guts).
● Any details (including tables of raw data and tables of calculated results) should be put in an
appendix.
● Remember to give units.

22
Conclusions

The problem of digital control of a delicate dynamic system has proven to be a difficult

but interesting challenge. The performance of our cube met the goals of the project. The cube is

able to withstand disturbances, return to a balancing state, and maintain this balancing state

indefinitely. If the project were to be done again, there are several things that could be done

differently. First, the absolute orientation “fusion” sensor would have been included in the

original design instead of attempts at using several IMUs to determine 3D orientation. Second,

the 1D prototype would have been replaced by the case of balancing on an edge of the cube. Too

much effort was put towards balancing a hastily made prototype when our final hardware

achieved the same test bench. Finally, the dynamics and control of balancing on an edge would

be started much earlier, and the consultation of an expert in the department would have been

arranged. These changes would have improved the success of our project, but ultimately the

overall goals of the project were achieved.

23
Appendices

· Should include all details.

· Should include CAD drawings and layout.
· Should include any raw measured data.

Appendix A: Previous Face Designs

The designs below were considered for mating the faces of the cube but were rejected

because they would not have provided a sufficiently clean edge or corner. Because the cube was

designed to balance on these features, it was critical that they be sharp and predictable. These

designs did not fulfill that specification.

Design 1: Symmetric Faces

Design 2: Overlapping Faces

24
25
Notes

· Please double-space all text. Please use 12-point type font.

· Page number amounts are for the written text and DO NOT include figures. Please make figures at least
1/3 to 1/2 page each so the details are clear. If your section exceeds these sizes, you are probably including
unnecessary information. However, make sure that you have all the information that I require even if the size
of the section is greater than the suggested page number amounts.

The final report must be written in English and submitted on 8.5”x11” paper, printed single sided and
presented at the Design Review. All pages must be numbered. All figures must be numbered and titled.

26