Selected Projects

for work, school, and others

Below is a portfolio of my design and design/build projects.  More design/build projects can be found on the My Projects page.  A number of my projects at Horizon Systems could not be shown publicly for confidentiality reasons. Questions welcome--

Plan view of an endtruck and portion of bridge beam (white), runway section (blue), and wheels (green).

Endview (cross-section) of endtruck at wheel centerline.  Simple wheel axles are supported in double shear.  Wheels are gear-driven by jackshaft and spider coupling (below, green) connected to gear-reducer (left, grey) (Motor and electric brake not shown). 

Bridge Crane: 2500 lb, 45' span, 70' runway
Horizon Systems, Inc.
2004, in progress
  • Responsible for all engineering, design,  installation procedures, and project management.
  • Designed/engineered in accordance with CMAA-74 Specifications.
  • Developed spreadsheet  to efficiently analyze multiple W+C bridge beam combinations; accounted for wheel contact + bending stresses and lateral torsional buckling.
  • Engineered/designed double-shear axle supports to reduce stresses and greatly simplify axle fabrication (endtrucks generally use cantlevered axles). 

Jib Crane: 2000 lb, 20' boom, Pole mount
Horizon Systems, Inc.
  • Engineered, designed, and supervised production.
  • Triangulated construction provides high stiffness--no bouncing loads.
  • Wide bearing span reduces internal forces and friction, increases stiffness.
  • Main beam analyzed for combined stresses due to wheel loads and bending. 
  • Friction minimized with cam-rollers for lower bearing and UHMW-PE for upper bearings
  • Shims in lower mount allow leveling of boom

Jib Crane: 2000 lb, 24' boom, Wall mount
Horizon Systems, Inc.
  • Engineered, designed, supervised production, supervised and assisted installation.
  • Similar to above crane, except bearings are  teflon-lined and bronze. 

Components for Rotary Airlock Feeder: Shaft/Bearing Assembly and Casting Alloy
Horizon Systems, Inc.

A rotary airlock feeds products such as grains or powders into a pressurized pneumatic conveyance stream, while minimizing air loss through airlock.  Deflections of the bearings and shaft, as well as thermal expansion of housing, increase clearances between rotor and housing.  Excessive clearances lead to excessive air and product leakage, which results in higher power requirements and poor conveying. 
  • Designed shaft and bearing assembly to minimize clearance and deflections due to pressure loads on rotor, yet allow thermal expansion.  Reduced bearing deflections 4x compared to competitors' airlocks.
  • Researched corrosion resistant casting alloys; selected alloy that provides lower thermal expansion, improved machinability, equal corrosion protection, and lower cost compared to 304/316. 

60,000 lb Baldwin load frame, SATEC hydraulics, and control panel.  

Close-up of the control panel, showing Newport meters and Yokogawa controllers.  LCD in upper right monitors the control output (milliamps) to the servovalve.  Labels not shown.  System has many control options:
  • Load Control for stiff specimens and the linear portion of a stress-strain curve.
  • Displacement Control for flexible specimens and the plastic portion of a stress-strain curve.
  • Manual/Continuous Control for a quick test of max load.
  • Manual/Incremental Control for slack removal and setting a preload.  I used a small DC generator (motor) for this and it turned out surprisingly effective. 
  • Control Combinations such as starting in Load Control then switching to Displacement Control for tensile testing of ductile specimens.

Instrumentation and Control System for 60,000 lb Baldwin/Satec Universal Test Machine
Valparaiso University, College of Engineering

While I was teaching at Valparaiso, the computer on their Baldwin/Satec servohydraulic testing machine went out.  The computer was too old to repair and the rest of the machine was in good shape, so I designed and developed new instrumentation based on commercial controllers:

Newport INFW Strain Gage Meter

Measures ram force and outputs to the load controller.

Newport INF8 Quadrature Counter

Measures ram displacement via the existing optical encoder and outputs to displacement controller.

(2) Yokogawa UP550 PID Controllers

One for displacement control, one for load control (switchable).  Controlling force on a universal test machine is considerably more difficult than a typical heat/cool application, due to both hydraulic and mechanical "slack."  Simplified:
  F = 0 for specimen strain<=0
  F = ~k*x for specimen strain>0
And, the controllers cannot know what ram displacement corresponds to strain=0.  (A heating analogy:  at some temperature unknown to the controller, the system changes from heating the Superdome to heating a closet.)  Since controllers have trouble with the stiffness discontinuity at x=0, I developed a "slack removal" profile segment to get through x=0 w/ little overshoot.  Once through x=0, I tuned 5 load-dependent PID parameter sets to optimize response.  (Autotune isn't very useful for this non linear system). 

These Yokogawas were the fastest sampling PID controllers I could find at a reasonable price, and while they aren't fast enough for dynamic tests, they run quasistatic profiles very well.

Experimental fixture to hold and sublux the shoulder and take stereo-xrays of the shoulder with calibration frame.  Xray source is above the picture, and rotates (with xray cassette) around the shoulder. 

Calibration frame and object used together to check accuracy of the SRG system.  The markers are 800 micron dia. tungsten carbide spheres.  3D accuracy is found by comparing two sets of 3D data for the Calibration Object markers: those found using a CMM (below, fitted with 300 micron ruby stylus), and those found using SRG.  Result: 20 micron RMS accuracy in 3 dimensions.

Coordinate measuring machine with video analysis system, to digitize the 2D coordinates of markers on x-rays.  The CMM is fitted with a macro-lensed video camera which focuses on a 1x1.5mm area of film. Video analysis (NIH Image macros) finds the center of each marker within that film area.  I choose this two-step process because it provides high linearity, accuracy, and repeatability (~5 microns) over the entire x-ray. 

Dissertation: Stereoradiogrammetry of Subluxed Shoulder Capsule
Orthopaedic Research Labs, Univ. of Michigan
Adviser: Lou Soslowsky, Ph.D.

My dissertation research received these awards:
  • Charles M. Neer Award for Excellence in Basic Science Research, American Society of Shoulder and Elbow Surgeons, 2000.
  • 1st place, Ph.D. Level Student Paper Competition, ASME Bioengineering Division, International Mechanical Engineering Congress & Exposition, 1998.
Significance: After someone dislocates their shoulder for the first time, they are often vulnerable to repeated  dislocations.  Surgeons believe this is because the shoulder ligaments/capsule become stretched, but no one had every shown this to be true and no one knew where or how much they were stretched.  I aimed to shed some light on those unknowns.

So, I wanted to measure the 2D strain fields of shoulder capsule/ligaments during a subluxation (a mini-dislocation).  I focused on the clinically most important antero-inferior capsule and discretized it into 45 elements.  Each element is only ~5mm on a side, so I needed to know the 3D node positions very accurately in order to find strain. 

I chose stereoradiogrammetry (SRG) to measure the node positions because 1) often the capsule is visibly obscured by the rotator cuff tendons and 2) it is highly accurate.  The basic idea of SRG is that if you have two views (xrays) of a 3D object, you can reconstruct its 3D dimensions, but, you generally need some reference points of known coordinates in both views.  These reference points are called "Calibration Markers" and are affixed to a "Calibration Frame" (left).

For high accuracy--and after first trying fused quartz and slip-cast porcelain--I choose carbon-fiber-epoxy for a dimensionally stable calibration frame.  I also made a calibration object of fused quartz.  I carefully epoxied 1mm dia. spherical markers to both of their surfaces and measured their 3D coordinates with a CMM.  Later, I fitted this same CMM with a super-macro-lensed video camera to digitize 2D marker positions on xrays.  Final 3D reconstructed accuracy:  20 microns--better than any prior SRG system--and good enough to calculate strain fields to +/- 0.5%.

To test the shoulder, I designed and built a fixture to position the scapula and humerus, apply simulated rotator cuff loads, sublux the humerus, and take stereo-xrays of the shoulder with calibration frame.  We applied a grid of small lead markers to the shoulder capsule to track the position of the capsule in both a nominal strain and subluxed state.  The images of the markers on the stereo-xrays were digitized (thank you Mike and Cameron) and 3D positions of the markers were reconstructed.  Strains fields were calculated from the position data using ABAQUS, and strain data was analyzed by Matlab and an SPSS model.

The results provided insight into how ligamentous capsule deforms, and how that deformation may recover after unloading.  That knowledge helps guide scientists in future research and surgeons in shoulder reconstructions.

Also see my Biomechanics page.

Ankle exerciser, shown with foot in inversion/eversion mode.  Spring rotates around lower left pivot to vary resistance; mechanism provides ~constant force for all spring positions. 

"1/x" function-generator: (Output angle) = 1 / (Input angle), and,
 (Output torque) = 1 / (Input torque)

Portable Ankle Exercise Device


As part of a graduate design course on product development, I led a group of seniors in their capstone design project.  A local company wanted us to design a physical therapy device for ankle rehabilitation.  It needed to be portable, low cost, and provide physiological resistance (relatively constant torque over the range of motion). 

My main design contribution was the resistance mechanism.  We used a pre-tensioned tension spring for low cost and weight, and transformed its linearly increasing force to ~constant force (torque) using a 4-bar linkage.  It's a fairly complex analysis, but here's the basic idea:

Spring:  Fspring = k * x
Mechanism:  Fout = Fspring *  (1/x)
Spring * Mechanism:  Fout = constant

I synthesized the "1/x"  function generator mechanism in Lincages.  It worked pretty well, especially having the pre-tensioned spring to give non-zero force right at 0+ displacement.

Shoulder Instability Experimental Fixture
Orthopaedic Research Labs, Univ. of Michigan

I designed and built this fixture for 3 experiments on the effects of muscle forces and shoulder ligaments on the stability of the glenohumeral joint.  The fixture positions the scapula and humerus and allows varying of 7 simulated muscle forces.  The entire fixture fits in the frame of an MTS machine, which subluxes the humerus.  The fixture fixes just enough humeral degrees of freedom to maintain the basic arm position, but frees all other DOFs to ensure a physiological subluxation.  The lines and pulleys are routed to 6 low friction pneumatic cylinders below the top plate.  Not shown is a 9-channel electropneumatic control box.  This fixture was used by other researchers for many years after my 3 studies with it. 

7-bar linkage schematic for load-sensitive camber control.  Light grey links represent a ~conventional double wishbone suspension.  Under cornering forces, black links transfer work from lower arm to upper arm, pulling upper arm in more than lower arm and thus producing negative camber.  The small spring-damper controls the rate of camber gain; it's absorbed work comes from a slight track change during cornering (F*d at the contact patch).  Conventional vertical spring/damper not shown. 

Load-Sensitive Camber Control Mechanism for a Vehicle Suspension


Finalist, Graduate Student Mechanism Design Competition, ASME Design Technical Conferences

I came up with this idea while working at Ford; at Michigan it evolved into a team project as part of a mechanism design course. 

Significance: Tires give the most cornering force when they have approximately negative 1 deg. camber (for outside wheel tire), but this is difficult to achieve without a trade off in braking, acceleration, and tire life, because conventional suspensions control camber via body roll (wheel travel).  Further, bushing deflections in all conventional suspensions actually produce more positive camber compliance than the negative camber they gain due to instant center geometry.  This universal problem struck me as an opportunity: could a suspension give "negative camber compliance"--i.e., negative camber due to cornering forces, independent of body roll.

The mechanism here does just that.  The principle is to transfer work from the lower arm to the upper arm; since lower arm forces are ~3x greater than upper arm forces, the upper arm is pulled in ~3x as much as the lower, producing negative camber.  There are many other ways to achieve this than a linkage, but a linkage is the simplest.  We synthesized the entire linkage to provide zero camber change in braking/accel and negative camber under cornering loads and body roll.

Our 2D prototype showed it does indeed work, but there are are some difficult 3D issues preventing production feasibility.

Completed and on the track.  Ours was typically the top finishing car that did not have factory support ($$).  

Rear frame, suspension, and powertrain during construction.   Spring/damper and antiroll bars not shown (these attach between the upright (black) and the center-upper hard point on the frame).  Rear bodywork attached as a unit.

IMSA GTU Porsche 911


For a few years after college I wanted to be a race car engineer.  So while both of us were working for Ford, Jay O'Connell and I teamed up with an IMSA GTU Porsche 911 owner (Jay Kjoller) to rebuild his car.  His existing car was a modified stock 911, with struts and semi-trailing arm suspensions.  The latter--in combination with the weight of the rear engine--is well known by 911 drivers for its "interesting" oversteer behavior. 

So we tore off the front and rear of his car and TIG-welded new triangulated structures with double wishbone suspensions--a first for a 911.  (Lesson 1: this is a lot of work, esp. while working a full time job.)  I concentrated on the rear chassis; Jay the front.  The rear suspension (left) is a 5-bar for ease of replacement in a crash, with roll and instant center control identical to a double wishbone.  We also made new bodywork and attempted a diffuser.  (Lesson 2: effective underbody aerodynamics requires a wind tunnel.) 

The car placed 4th in its first race at Mosport--and was the top finishing privateer (GTU was made up of ~$1M factory teams and ~$50k private teams).  Throughout the season we were usually the top placing private entry.  One of the hired drivers said it was the best handling 911 he had ever driven.  Jay kept with the project after I went to grad school, and years later headed engineering for Ford/Jaguar's F1 program, among many other accomplishments.  

1st place overall at the '88 competition.  The main barrier to a stiff racecar frame is the cockpit opening.  Frame designers want to diagonalize that "open face", but drivers tend to be disagreeable about a steel tube running through their ribcage.  We fixed the problem with "pyramids" on either side of the open face. 

Front frame and suspension during construction.  Tubes are 1 x .028"  4130 steel; joints are TIG welded. 

Cornell Univ. Formula-SAE Competition Racecar


I joined Cornell's FSAE team for my senior year, and like many FSAE-ers became immersed in how much fun engineering can be. 

I focused on frame design/construction and the rear differential (a Torsen).  At the time we didn't trust our skills with FEA programs, since the previous year's team used one but our actual tests showed it had a torsionally floppy frame.  So we learned about space frames the old fashioned way--triangulation theory and scale models.  As a beginner's project, I think we both learned more and had a better design without FEA:  we were forced to understand the fundamentals, and from that understanding we rapidly converged to a sound design.  Had we had FEA, we would probably have fallen into the trap of iterating a mediocre design, which usually results in more mediocrity.  Of course, iterating on a sound design gives an optimal design.  Years later at Michigan I advised their FSAE team and did some of that FEA optimizing using I-DEAS on generic models--but not before teaching them the fundamentals with wire frame physical models.   The team came up with clever and sound solutions and then optimized them using FEA. 

In retrospect we triangulated the '88 car in excess, as the car was not mass-produceable.  And it takes a very long time to TIG weld all those .028" wall joints.  But we did achieve 3200 ft-lb/deg at only 33 lb frame weight, which is respectable even today.

The big lessons from the FSAE project were 1) creative mechanical design is tremendous fun to me and 2) exchange and development of ideas in a team setting is highly rewarding and results in a better design.

Also see the My Projects page.