A How-To Guide to Reverse Engineering

June 3, 2019
3D scanning and CT scanning have extended the science of metrology into the realm of design by converting an object into a CAD model.

Taking a lead from a product designer is just the most obvious way that a foundry develops a production plan. It’s equally possible to start with an existing part and work back toward a functional design that can be the basis of a production plan.

This is possible because 3D scanning and CT scanning have extended the science of metrology into the realm of design by converting objects into a CAD models. So, where to begin?

Step 1: Determine intent
“How will this image be used? Can we anticipate any changes? What are the tolerance requirements?” All these questions are paramount for determining the successful path of the data output — and each one is mutually exclusive of the others.

Why does it matter how a 3D image will be used? Because for any project the proper approach is to find the clearest path to an optimal result. Not knowing the design output can send a 3D service tech down a very long, inefficient path.

Here is a scenario I encounter on a regular basis. Customer: “I need to scan an entire engine for a vehicle, with CAD output.” Shall we jump to the conclusion that he needs every nook and cranny digitally captured by any medium, and that we can spend weeks meticulously creating detailed models of starter motors, cooling lines and complex engine block castings? That’s almost never necessary. Typically, a simple volumetric representation is all that is needed to determine fit and clearances. Simplified scanning techniques and rapid solid modeling can accomplish this quickly and efficiently.

As for anticipated changes, these can drive the method of creation. Also commonly heard from customers is, “We need a Parametric IGES or STEP file.” This is a contradiction. Parametric CAD models contain design intent, a structured combination of prismatic 3D features driven by specific dimensions. An IGES or STEP output is a stripped version of that parametric 3D CAD model, typically represented by that outside “skin” of the CAD surface, allowing users to share 3D data between multiple platforms that do not share a common internal language.

Knowing what changes may be needed, a user can design a part to make those anticipated areas easier to modify without fighting unwieldy surfaces, fillets, or draft in the completed CAD model.

If there is no plan to modify the data — as with the engine example, only to check what fits around it — this can be resolved with a simplified NURBS (Non-Uniform Rational B-Splines) to get a lightweight, volumetric skin to bring in to the customer’s existing CAD design.

The final deliverable should be perfect, right? Wrong. Do you really need this sand casting within 0.001 in.? Are you sure you want to see every imperfection in that part? Let’s back up a bit.

Reverse engineering is a path with many twists and forks. Tolerances can be managed by a skilled technician using the right hardware and software. Expectations are more difficult to manage.

Modern scanning hardware possesses the ability to capture high surface detail, sometimes to the detriment or advantage of the process, depending on the factors mentioned above. The reason is the anticipated changes.

If the intent is to create new tooling and develop clean surfaces for machining, then developing it through a traditional CAD workflow makes the most sense. But, this calls for an allowance of “tolerance.” That intended flat surface may not be flat on the physical part. By 3D scanning and interpreting those surfaces back to a clean CAD model, correcting that discrepancy influences the deviation on the actual part back to the new digital CAD counterpart.

On the flip side, if I am developing something to perhaps grab or hold this part and I need high precision, then back to that prior fork in the road. We can now take a different approach to the modeling output. Similar to the engine example, we can create high precision NURBS output to satisfy those exact surfaces and make certain the part is defined properly.

Step 2: Acquisition
Having established the guidelines for intent, let’s examine our options, based on those choices. Structured light is cleaner, portable arm-based laser scanning is much faster and more accurate, plus the time-of-flight and phase shift (long range) scanners can scan further distances with substantially higher precision. Metrology grade 3D CT (X-ray) scanners are becoming more powerful and financially feasible.

Why structured light? Because clean data yields a cleaner result. Structured light is typically a two-camera, stereo system. A fringe pattern is digitally projected onto the surface of the part, thus displacement of the fringe pattern along the part is correlated back into 3D data. These sheets of lights bounced off the part provide a clean and highly accurate digital representation of the part. This clarity is regarded as a high standard in comparison to its counterparts (mentioned below.) Its only real limitations are translucency/transparency and deep colors opposing the light spectrum of the projected light. Also, both cameras need to see the geometry that is being captured.

Why portable CMM/scanning arms? Speed and flexibility. Portable CMM’s wider laser lines, higher hertz rates for data capture, and millions of points captured in seconds allow users to resolve issues quickly, from a controlled lab environment to a shop-floor setup.

Data from these units are captured via laser and its ability to adapt to different surface colors and finishes is now highly advanced. Current limitations are set only by the length of the portable arm, with multiple set-ups required for part sizes beyond the arm’s reach.

Why long range? Do you need to map out a building? Do you need to reverse engineer the outside of a 747? If so, this is the perfect tool. With the ability to scan geometry hundreds of meters away within a reasonable tolerance, long-range scanning is the right application. These tools send out a laser beam with high precision and record the surface it bounces off back into digital 3D data. That data can be combined with high-res imagery to provide 3D visualization of the objects, areas or spaces being scanned.

Why CT (X-ray)? Visualizing internal data is typically categorized as inspection. But with 3D X-Ray machines now able to see into dense materials (e.g., steel), extracting those internal images can prove valuable. Internal passages that were created with a complex network of sand cores can be seen to correctly validate clearances when recreating a new part. CT also eliminates blind spots, allowing designers to model complex sculptural surfaces with precision and assumptions in filling in missing geometry from conventional scanning methods.

Step 3: Processing
Now we know what the intention is and how we are acquiring data, so what do we do with it? Today’s processing is a a broad spectrum, but here are guidelines to lead us down a clear path.

a) Garbage in, garbage out. With the variety of scanners mentioned above comes a bit of responsibility. Know your hardware; know its capabilities and its limitations. We’ve learned how it’s being used, how accurate it needs to be, and what methods we need to use to get there. Taking small short cuts in the scanning process leads to time-consuming editing when processing 3D data.

For instance, not taking an additional scan to capture the bottom of a groove or hole leads to a lot of assumptions when interpreting the scan data into a polygonal mesh. One more five-second scan could save hours of work. Clean data input streamlines the processing.

b) 3D point cloud, now what? We have millions of points that cleanly represent the part, and the common next step is to convert the XYZ 3D data points into a polygon model. Simply put, the software connects the dots with a series of triangles to create a representative skin.

There are various tools that can accomplish this goal. Most hardware suppliers provide this direct output from their scanners, others rely on third party software to run the calculations. From here, there are many software packages for manipulating the data, including smoothing out imperfections, closing small or large holes within a reasonable assumed precision. This sets the pace for the next round of modeling.

c) Conversion. Now the polygon data, commonly referred to as an STL, is ready to go. Intent? Check! Now let’s convert this to a useable format with the necessary parameters to get what we need out of it. This is final fork in the road, as defined earlier by our intent. It’s another part of the process that has dramatically improved in the last decade. The ability to get from physical part to 3D data is 5:1, maybe more. Gone are the days of scanning a part, bringing a low-resolution version into a CAD package, cutting cross sections, converting those to complex sketches, and then generating 3D features from those sketches.

Today, there are software packages that can handle scanning directly into software, converting large data sets to high-resolution mesh data, and generating native parametric CAD features, all in one package, thereby shaving days off the process and achieving a much better result.

Also, the ability to generate NURBS or “as-is” surface data with extremely high precision has been a dramatic improvement. With complex algorithms to solve data sets with complex surface geometry at the push of a button, the process continues to get faster and faster, as well as more accurate.

Validation. Now that we have 3D scan data and the intended CAD model, let’s wrap up the process. Before I pass the results to manufacturing (or another downstream process), I need to check my work. Another improvement that continues to improve is the ability to validate data. Validation, used here, is the ability to show deviation of the scanned object back to the CAD model being developed. This deviation is represented typically by a color map, with each color representing the 3D distance each point varies from its CAD model counterpart.

Once this evaluation is complete, and meets the expectations determined, the CAD model is ready to be delivered. 
Greg Groth is the Midwest services manager for Exact Metrology offering comprehensive metrology services, including 3D laser and CT scanning, reverse engineering, quality inspection, product development, 3D printing, and 2D drawings. Contact him via LinkedIn.