remade from scratch
A military-sponsored project mounts a collaborative effort to reverse engineer legacy parts.
The U.S. military owns and operates many
complex electromech- anical systems that were designed 25 to 50 years
ago. Because of the cost of replacement, these systems may have to be
used for decades to come, well beyond their intended design life.
Maintenance requires spare parts, but in many cases, the original manufacturers
are no longer around to provide them. So the military needs a comprehensive
plan to determine how best to prolong the life of these legacy systems
and, in some cases, new technologies to reverse engineer critical parts.
"We need a holistic strategy to research in legacy systems engineering,
not just clever point solutions," said David Hislop, head of the
Army Research Office's Software and Knowledge-Based Systems program.
Manufacturing parts for old systems can be difficult because documentation
about the components may be unavailable, incomplete, or in a form that's
incompatible with modern computer-aided design and manufacturing software.
The legacy component has to be engineered to interface properly with other
remaining parts of a system. The engineering strategy also has to consider
how to upgrade the part to take advantage of materials, manufacturing
methods, and analysis tools that may have improved since the parts were
originally designed. And there's the consideration of whether to reverse
engineer, re-engineer, or completely redesign the part.
Deciding which approach to use requires not only a technical assessment,
but also an economic, cost-benefit analysis. In such assessments, time
may be an overriding factor.
The military created the Virtual Parts Engineering Research Initiative,
or VPERI, to provide the vision, strategy, and engineering to help solve
its legacy systems problem. The program, funded by the Army Research Office,
is a collaborative effort for building frameworks, tools, and technologies
for making engineered systems sustainable and maintainable in the 21st
century.
This virtual engineering environment is intended to transform the engineering
process, to provide extremely fast turnaround times for urgent part supply
needs. The Virtual Parts Engineering Research Initiative is in its second
year of operation. Current participants include Hampton University in
Hampton, Va., the University of Utah in Salt Lake City, Arizona State
University in Tempe, and the South Carolina Research Authority of North
Charleston.
Recreating a Gearbox
In order to test the analysis tools developed under VPERI, the collaborative
team set out to reverse and re-engineer a legacy system, a gearbox from
Northrop Grumman Newport News. The gear was originally used to drive a
crane at the shipyard in Virginia.
This gearbox was chosen as a real-world reverse-engineering project because
it exhibited several characteristics that made it a good test of the analysis
tools. It required multiple manufacturing processes and involved partial
redesign. It was highly modular with simple, clean, and well-defined interfaces
to the outside, and the number of internal parts was relatively small.
The original gearbox uses two-stage reduction with helical gears, has
a reduction ratio of 38.9 and a horsepower rating of 4.5 at 1,750 rpm.
The gears were non-standard. No drawings, CAD models, or design calculations
were available.
At Hampton University, Wallace Arnold, director of the Data Conversion
and Management Lab, and Vadivel Jagasivamani, an adjunct research professor,
extracted part and assembly information using manual measurement, photography,
and ultrasonic imaging, all without disassembling the gear.
 |
| The original gearbox that was to be re-engineered used to drive a shipyard crane. |
Elaine Cohen and Richard Riesenfeld, co-chairs of the University of Utah's
Geometric Design and Computation Group, used state-of-the-art data acquisition
techniques, including ladar scanning (laser radar, a 3-D spatial measurement
tool), to produce a point cloud sampling of the surface of the part.
This data was fitted using feature-based and freeform surface algorithms
developed through VPERI research.
Instead of attempting to manufacture an exact replica of the original
gearbox, the part was re-engineered and improved by using knowledge of
the latest manufacturing capabilities, materials, and analysis tools.
Arizona State's Design Automation Lab, under the direction of Jami Shah,
used a knowledge-based parametric design shell to demonstrate redesign
and engineering analysis. Sam Drake, a research associate professor of
mechanical engineering and computer sciences at the University of Utah,
manufactured the end product, a re-engineered gearbox.
The Virtual Parts Engineering Research Initiative also considered completely
redesigning the gearbox, as well as the possibility of replacing it with
a commercially available part. Although both of these approaches would
be cheaper and quicker to implement, the team decided against them since
the gearbox made a good test case for collaborative reverse-engineering,
a process the military sees as vital in lengthening the useful life of
weapon systems.
Both the re-engineering and redesign approaches required the team to come
up with functional specifications and interfacing requirements. The parametric
design shell from Arizona State's Design Automation Lab was used to demonstrate
the archiving of machine-design knowledge and design retrieval.
The shell lets designers add knowledge about different applications in
terms of key parameters and their relationships, which represent the functional
behavior of the components and physical laws that govern the behavior,
spatial arrangements, and other attributes.
After this knowledge is added, designers use the shell to define the particular
design problem in terms of these variables and design objectives. Then,
the shell provides mechanisms that enable designers to ascertain that
their design has met the functional requirements and helps them explore
alternatives.
Using the Arizona State design shell, it took eight hours to build the
knowledge base. Exploring the redesign of each new gearbox took about
two hours. At the end, the redesigned gearbox specifications were produced,
complete with assembly and part CAD models, and a bill of materials that
included commercially available standard components.
While the entire gearbox was reverse-engineered on paper, only the housing
was manufactured from scratch at the University of Utah. The rest of the
gearbox was put together using commercial parts.
Separate, but in Touch
All three partners participated in the reverse-engineering process through
a coherent, seamless flow of data, information, and artifacts. The geographic
separation of this widely dispersed virtual design team didn't pose a
problem.
Input to the reverse-engineering process came about in any of three ways—namely,
scanning a physical artifact with lasers or other applicable technology,
analyzing drawings for geometry and annotations, and performing a variety
of engineering analyses to determine appropriate re-engineering specifications
in terms of current materials and processes.
Scanning produces point cloud data, which in turn feeds both feature extraction
and surface extraction modules to produce a rough model of the part that's
to be built.
The rough model is used along with feature knowledge to drive a coordinate
measuring machine so that data can be refined. The CMM is a mechanical
system that moves a measuring probe to determine coordinates of points
on a surface.
 |
| While the whole gearbox was reverse engineered on paper, only the housing was manufactured from scratch. |
The CMM uses a probe to determine the depth of a hole, for instance,
or whether, in fact, it is a through hole or a blind hole. This information
might not be provided with certainty by scanning technology.
Legacy drawings can be scanned and analyzed, as is data abstracted from
the available legacy system. Both of these analyses use different software
modules, providing input and materials properties calculation and specification.
The drawing analysis can also feed into methods for extracting surfaces
from line drawings, to provide another source of information for inferring
a rough model. Similarly, if material properties can be extracted or calculated,
they provide additional information to help create the refined model.
Once the high-level feature-based model is validated with accurate coordinate
measurements, the re-engineering process enters into the formal definition
phase. During this phase, STEP modeling language CAD drawings are generated
for archiving and procurement. These refined drawings can then be used
to drive the in-house manufacturing process, or they can be sent out to
bid for procurement of the replacement parts.
Specialized Requirements
Legacy system engineering is typified by small batch sizes, short delivery
times, high product variety, and incomplete information about design history,
rationale, and specifications—all of which lend it different requirements
and priorities than those in play in a conventional product development
process. Besides requirements, these factors also affect materials selection
and manufacturing methods. A high degree of automation of design, simulation,
and manufacturing planning, as well as task interoperability, is even
more important for legacy system engineering than it is for conventional
product development.
Maintaining significant legacy data systems and the process of reverse
engineering are intimately connected. Maintaining an important legacy
system entails producing a good, contemporary replacement component based
largely on data extracted from a failed or outdated component.
In some cases, the designer must deduce the function of the undocumented,
nonfunctioning artifact. Then there's the puzzle of figuring out
the intended purpose of some aspect of a design that appears to play no
meaningful role in the way the designer imagines the system to perform.
For a family of components, an enigmatic feature may be vestigial in character,
or it may be present because of the function played in a sister part for
some family of components. All of these characteristics make legacy system
engineering necessarily interpretive and occasionally speculative.
Editor's Note: This story is based on research from
the Virtual Parts Engineering Research Initiative team. Wallace Arnold,
Elaine Cohen, Tom Henderson, Vadivel Jagasivamani, Rich Riesenfeld, and
Jami J. Shah all contributed to the piece.