Two types of engineering:

1) Forward engineering:
High level abstractions (means defining system’s design, components and interaction on
conceptual level) and logical design to physical implementation of a system.
2) Reverse engineering:
It is process of duplication existing part, subassembly or product without drawings,
documentation or computer model.
Also defined as process of obtaining CAD model from 3D points (cloud points) obtained by
scanning existing parts/products
Application of reverse engineering: Manufacturing, industrial design and jewelry design and
reproduction
Advantage: Using reverse engineering we can quickly capture a physical product in a digital
form, remodel it and export it for rapid prototyping using 3D printer or CNC machine. This
shorten product development time
 RE tech. is used for reconstructing 3D models of object in different geometric formats
 RE tech. are of 3 types: contact, noncontact and destructive
 Output of RE’s data acquisition process is 2D images and point cloud representing geo.
of object
 RE software transforms data from RE hardware into 3D model
 Final output of data processing is 2 types of 3D data: 1) Polygon 2)NURBS (non uniform
rational B-splines)
 Polygon used for: rapid prototyping, laser milling, 3D graphics, simulation and
animations.
 NURBS used in: CAD,CAM and CAE
Q) How RE reduce product development time (CHAT GPT)
A) Reverse engineering can indeed contribute to reducing product development time for
mechanical engineers in several ways:

1. Quick Understanding of Existing Designs: Reverse engineering allows mechanical engineers to

quickly understand the design and structure of existing products or components. This
understanding can serve as a solid starting point for designing new products, saving time that
would have been spent on creating a design from scratch.

2. Rapid Prototyping: By reverse engineering a prototype or an existing product, engineers can

create accurate CAD models that can be used for rapid prototyping using techniques like 3D
printing or CNC machining. This accelerates the prototyping phase and helps engineers iterate
and test their designs more quickly.

3. Design Validation and Improvement: When reverse engineering an existing product,

engineers can identify design flaws or areas for improvement early in the development process.
This eliminates the need for multiple design iterations during the later stages, saving significant
time.
4. Reusing and Modifying Existing Designs: Reverse engineering allows engineers to reuse or
modify existing designs, components, or parts. This can be particularly valuable when designing
variations of a product or adapting an existing design for a new application.

5. Fast Customization: When designing custom parts or components, reverse engineering

enables engineers to quickly tailor designs to specific requirements. This is especially useful
when off-the-shelf solutions are not suitable.

6. Faster Replacement Part Development: In cases where replacement parts are needed for
maintenance or repair, reverse engineering can expedite the process. Engineers can quickly
recreate the required part without starting from scratch.

7. Accelerated CAD Modeling: Reverse engineering provides accurate measurements and

dimensions that can be directly used to create CAD models. This eliminates the need for manual
measurements and speeds up the modeling process.

8. Reduced Trial and Error: With a better understanding of existing designs obtained through
reverse engineering, engineers can reduce the trial-and-error phase during product
development. This leads to more efficient decision-making and design choices.

9. Streamlined Documentation: Reverse engineering can help in creating accurate

documentation for existing products, which can be valuable when integrating those products
into larger systems or projects.

10. Minimized Research Time: Reverse engineering can provide insights into competitors'
products or similar solutions in the market. This can help engineers quickly understand industry
trends and best practices, reducing the time spent on research.

11. Enhanced Collaboration: Reverse engineering allows engineers to share a tangible

representation of the existing product's design. This can lead to better collaboration among
team members and stakeholders, leading to faster decision-making.

Reverse Engineering (generic process) : 3 phases:

1) Scanning
2) Point processing
 3) Application-specific geometric model development

Considerations:
1) Reason for RE
2) Number of parts to be scanned
3) Part Size
4) Part Complexity
5) Part material
6) Part finish
7) Part geometry
8) Accuracy needed

Phase 1 – Scanning: Choose correct scanning technique, prepare part to scan, perform scanning to
capture info. describing all geo. features of part (slots, pockets, holes)

Scanning devices can be a dedicated tool or add-on to existing computer numerically controlled (CNC)
machine tools

2 types of scanners: contact type and non-contact type

CONTACT SCANNERS:

 Contact probe follow contour of physical surface

 Based on CMM technologies, tolerance: +0.01 to 0.02mm
 Contact method is slow
 The probe must deflect to register point, so contact pressure is maintained. This limits the use of
device for soft materials like rubber which cannot be easily or accurately scanned
NON-CONTACT SCANNERS:

 A variety of noncontact scanning tech available.

 Noncontact devices uses: lasers, optics and charge-coupled devices(CCD: It converts incoming
light (photons) into an electrical charge (electrons), which can then be processed to
create digital images) sensors
 Collect data faster

Issues:

 Tolerance within ± 0.025 to 0.2 mm

 Some technique has problem generating data for (surface) || to (axis of laser)

 Problem when laser strikes shiny surfaces, so we need to prepare part using temporary coating
of fine powder b4 scanning

Accuracy matters more than speed of scanning => restricted use of remote sensing devices

Due to R&D in optical tech., accuracy of noncontact scanners is improving

PHASE 2: POINT PROCESSING:

 It involves importing cloud points, reducing noise in collected data and reduce number of
pts..These task performed by predefined filters
 Need understanding of filter algo so they know which filter to use for each task
 Processing allows to merge multiple scan data sets
 Part is scanned multiple times to ensure each feature is scanned (by rotating part) => each
scanned data is crucial. Good datum (Datum planning involves selecting reference points or
planes on an object's surface that serve as a coordinate system for aligning and merging
multiple scans) planning for multiple scanning reduces effort required in point processing
 Output of point processing is clean, merged, point cloud data set in most convenient format

PHASE 3: APPLICATION GEOMETRIC MODEL DEVELOPMENT

Current RE tech., helping to reduce time to create electronic CAD model from existing physical
representations
  Surface fitting algo are needed to generate surface that accurately represent 3D info. described
within point cloud data sets
 Most CAD software cannot display and process large amount of point data, so RE modules or
discrete software packages are needed for pt. processing
 This phase depends on real purpose for reverse engineering
 RE software can also be used for inspection by importing CAD model used for designing a
product and superimposing scanned point cloud data on it

REVERSE ENGINEERING HARDWARE:

1) Contact Methods
 They use sensing devices with mechanical arms, coordinate measurement machines
(CMM) and computer numerical control (CNC) machines, to digitalize surface
 2 types of data collection tech.:

Point-to point sensing with touch trigger probe:

 Touch trigger probe installed on CMM or articulated mechanical arm
 Manually operated: articulated mech. arm + touch trigger probe – allow multiple DOF
 CMM + touch trigger probe can be programmed to follow planned path
 CMM gives accurate measurement data compared to articulated arm
 Limitation: CMM lack no. of DOF(cannot digitalize complex surfaces in same way as
articulated arm)
Analogue sensing with scanning probes:
 Scanning probe – installed on CMM or CNC machine
 The probe provides continuous deflection output which combines with machine pos. to
get location of surface
 While scanning, probe stylus tip contacts the feature then move along surface
(gathering data as it moves). So we need to keep deflection of probe withing
measurment range of probe
  Scanning speed in this method is 3x faster in point-to-point sensing

ADVANTAGE OF CONTACT METHOD:

 High accuracy
 Low cost
 Ability to measure deep slots and pockets
 Insensitivity to color or transparency

DISADVANTAGE OF CONTACT METHOD:

 Slow data collection
 Distortion of soft object by probe.

2) Noncontact method:
 2D images and point cloud representing geo. of object are captured by projecting
energy sources (light, sound or magnetic field) onto object.
 Transmitted or reflected energy is observed
 Geo. data finally calc.ed using triangulation, time of flight, wave interference info. and
image processing algo.
 No contact btw RE hardware and an obj during data acquisition
 Non-contact RE can be classified on basis of: sensor tech., data acquisition technique

ADVANTAGES OF NONCONTACT METHOD:

 No physical contact
 Good accuracy and resolution
 Fast digitizing of substantial volumes
 Ability to detect color
 Ability to scan highly detailed obj, where mech. touch probe may be too large to
accomplish task
 DISADVANTAGE OF NONCONTACT METHOD:
 Limitations for colored, transparent or reflective surfaces
 Lower accuracy

OPTICAL TECH.

Laser scanner use geo. triangulation to determine surface coordinate

Triangulation method: use location and angles btw light sources and photosensitive devices (CCD-charge
coupled devices) to calc. coordinates

SINGLE CAMERA SETUP:

 Device transmits light spot on obj at defined angle.

 CCD detects pos. of reflected point on the surface

DOUBLE CAMERA SETUP:

 2 cameras are used

 Light projector is not used in measuring function and may consist of moving light spot or line or
pattern

PRINCIPLE OF TRIANGULATION METHOD:

 High E light source – focused and projected at specific angle (Q) onto surface of obj.
 Photosensitive device sense reflection from illuminated point on surface
 Baseline length(L) / Calibration: fixed distance btw light source and camera
 Angle (Q)
 Focal length of camera(F)
 Image coordinate point of illuminated point (P)
 Pos. of illuminated point (Pi) wrt camera coordinate sys given by:

 Increasing base line length => higher accuracy but due to practical reasons we can’t inc. base
line length, limited by hardware structure of scanner
 So, used for smaller obj over short distance
 If single-pt. or sheet light pattern used as light source, scanner mounted on travel platform, to
prod. multi. surface scans

STRUCTURED LIGHT:
  A light pattern is projected at a known angle onto surface and an image of reflected pattern,
reflected by surface is captured
 Image is analyzed to calc. coordinates of data points on surface
 Light pattern: single point, sheet of light, strip, grid,etc
 Most common pattern : sheet of light
 When sheet of light intersect obj, line of light is formed along contour of obj
 Line is detected and X,Y,Z coordinates of pts. Along line are calc. by triangulation
 Sheet of light sweeps obj as linear slide(carrying source of light) moves in X dir while a sequence
of img. are taken by camera
 Index number k assigned to each img. taken in order
 So k correspond to X pos. of sheet of light
 Image k got image coordinates (i, j)
 (I,j,k) are range image coordinates which are transformed into (x,y,z) world coordinates using
calibration matix

ADVANTAGES of structured light over laser systems:

1) Data acquisition is very fast

2) Color texture info is available
3) It does not use laser
4) Above features allowed digitalization of images of human beings

DESTRUCTIVE METHOD:

 Used for small and complex obj. in which both internal and external features are scanned.
 CNC milling machine expose 2D cross sectional img. which are gathered by CCD camera
 In RP process, part-built layer by layer using 2D img. data
 RE is reverse of RP process, 2D data is gathered by destroying part layer by layer

DISADVANTAGE:

 Loss of Original Product

 Costly process

ADVANTAGES:

 Fast
 Acceptable accuracy
 Repeatability is ± 0.0127mm
 Layer thickness: 0.0127 – 0.254 mm
 Allows capturing of internal features
 Can be used on any machinable obj.

WHAT IS CARE? :

 CAE(using CAD and CAM tech.) is automation of engineering an fabrication where a design
formalizes ideas through computer modeling and then fabricate those models into real world
 objects. CAD softwares are used to create digital models and CAM softwares are used to give
instruction to machine and tools for converting these digital models into real life obj.
 CARE is opposite of this
 It creates digital models of obj through measurements of obj.
 First step in CARE is making measurements at pts. along surface of obj.
 Each pt. has x,y,z coordinates in 3D space
 Collection of these pts. is pt. cloud. It gives sufficient desc. of obj
 Higher level possible by using more surface features such as surface edges and creases from pt.
cloud data.

ADVANTAGES OF CARE:

 Its allows rapid inspection and validation of real life obj. based on original CAD designs
 Can quickly evaluate tolerance relative to CAD model
 Such feedback enables tighter control over manufacturing process or aids redesign of
component
 It allows faster and easier transfer of data as compared to traditional method of expensive
courier delivery of physical prototype

HOW TO AUTOMATE measuring of geo. of obj? : CMMs and Computer vision systems

Coordinate measuring machine (CMM):

 Vernier Caliper- manually measure objects, requires significant effort for complex obj.
 CMM tech. – first step taken to automate RE process
 CMM still used and popular for simple tasks
 For simple obj., manual process of measurement using caliper is straightforward but as
complexity of obj. inc. , primitive planar and quadratic surfaces are no longer suitable
 To solve this prob, CMM was introduced.
 It has 3 mutually perpendicular axes (X,Y,Z) each has built in reference standard

 Probe allows accurate measurement along each axis relative to standard

  Operator can run CMM in manual mode or may be program the probe to maneuver
automatically
 CMM has features which help operator to plan the probe path
 Advance CMM sys allow operator to upload CAD model of obj and then use it for path planning
strategy
 CMM analyze CAD model to identify critical pts. and regions such as discontinuity or tight
spaces
 Tight space are emphasized cuz probe must come in contact with point or be close to the point
being measured
 This limitation lead to development of computer vision

COMPUTER VISION:

 Uses imaging-based sensors such as video cameras

 Stereo vision: 2 cameras operate in a manner similar to human eyes
 It has become useful for navigation planning and obstacle avoidance for mars rover missions
 It lacks accuracy that RE application demands
 Limitation: establishment of correspondence btw one pt in one camera’s img and same pt. in
second camera’s img
 Passive stereo vision: uses ambient light, no active forms of illumination used
 One way to overcome correspondence problem is to control illumination, therefore active
illumination stereo vision is used.
 In it, we replace one stereo camera with light source (usually laser).
 It allows more accurate depth measurements
Active Stereo (AS) example of continuous wave system

 Laser travels as wave from source to obj. and then back to detector
 As a wave, light undergoes phase change
 Difference in initial phase and final phase use to find obj’s distance

AS, example of Time of Flight system

 Pt. laser emit pulse onto obj.

 Difference btw. Initial time(wen pulse was transmitted) and final time(wen it returns to
detector) used to find obj’s distance

AS, example of Structured Light system:

 B, α, β are known quantities due to system calibration.

Benefits and Drawback:

 Theoretical comparison:
 CMM are mechanical sys, bound by physical principles i.e. probe must physically maneuver
around obj.
 Camera (are solid state sensors) take measurements from a distance and require less
mechanical precision
 Accuracy of CMM depend on mechanical control of probe
 Accuracy of active stereo sys depend on solid state sensors

Advantage of standoff nature of computer vision (i.e. camera is at some standoff dist. Away frm
obj):
 Avoid damaging to obj and probe: as probe travels on surface it may scratch it, if the temp. of
surface is too hot or too cold then heat transfer can dmg probe
 Simplification of path planning: in CMM- need to plan path carefully to avoid tight spaces and
discontinuity but there is no such limitation for computer vision

Source Illumination Categories:

 Sheet of light is faster than point laser cuz point laser need to scan MxN pixels but sheet light
only need M pixels for the same size of image i.e. sheet light will cover the whole geo. of part
line by line but point laser will cover whole geo. point by point.
 Primary difference btw them is speed with which they collect data, accuracy is almost the same
for all method
 Coded light method has a bottleneck on mechanism needed to display patterns on object
 Sheet of light is fastest

Sheet of light range imaging

 Laser projects line on obj. and triangulation among line, laser and camera give 3D
measurement
  Convery belt can used to move obj. under laser line to scan profiles over the obj.
 Another approach is not move the obj. but the scanner.
The laser and camera are mounted on wand. User swipes wand arund obj.

Scanner Pipeline (it is scanning of obj + making digital model):

1. Data collection:
 Sheet of light is best suited for CARE application
 1st step (after data collection) Calibration allows transforming of range image into 3D

pt. cloud
 Range img. is similar to conventional digital img. (one we find in digital cam)
 Calibration determine relationship btw img. coordinates (r,i,j) (r is range value units are
relative to config. of sys.) and world coordinate system(x,y,z)
 Left img in fig. 2.13 is range img.
 Right img is point cloud, diff. to visualize on 2D sheet but can be easily viewed in a 3D
viewer
 2nd step View registration: Fig. 2.14 shows u need at least 2 views to collect
measurements from entire part
 Challenge: we need to define 2 coordinate sys.
 Consider pt. P in one view img. has (x,y,z) coordinates relative to one coordinate system
but this same point will have p’ = (x’,y’,z’) coordinates relative to the second coordinate
system on the other view img.
 P and P’ are physically same pt. so we will need to apply transformation, eq. of

transformation:
 R is rotational matrix
 T is translational matrix
 The problem is that R and t are unknown
 Registered algo. find R and t using raw data from diff view of obj
 If no overlap exist in the view imgs then registration is almost impossible, we will need
higher order constraints
  3rd step is Mesh Reconstruction:
 After making a point cloud we need to mesh the model
 Triangle mesh are default standard as they simply computer visualization cuz they have
guaranteed convexity
 Sampling: defines no. of pts. in pt cloud and their relative distance to each other
 Under-sampling: incomplete and inaccurate model
 Over-sampling: overwhelm computer resources.
 Mesh reconstruction algo must define their sampling requirements
 Measurement error: If we have 3D point p on surface of obj, measuring it with scanner
will have some errors. If the observed pt is q then relation btw the pt is:

vector e is error
 Mesh reconstruction tries to minimize the impact of e
 Mesh reconstruction algo must consider 3 factors:
1) Topology of obj.
2) Overlap btw diff view of obj.:
 primarily for registration but it also useful to minimize measurement error
3) Missing data:
 It leads to holes in mesh

2 categories of mesh reconstruction algo:

1) Surface based
2) Volumetric based

4th step Surface fitting:

 Once we have first order approximation of obj from mesh reconstruction, final stage is
higher order desc. that is needed for CAD applications
 In some cases triangle mesh from reconstruction are sufficient
 In other cases we need parametric rep. such as nonuniform rational B-splines(NURBS)
 Last stage is broadly labeled as surface fitting
 Parametric rep. of surface:

 U and v are parameters defining surface

 F,g,h functions return x,y,z coordinates
 One surface is not enough for a complex obj.
 We need to divide surface into different surface patches with their own parameterized
surface
 Major challenge: identify appropriate patch cuz parametrized surface assume rectangle
structure which does not fit arbitrary boundaries that occur with path selection.