Scanning systems and AI-supported CT analysis software provide the German automobile manufacturer BMW with crucial information for producing housings for electric motors of their IX model EVs.

The moving assembly line has proven its worth since Henry Ford introduced the industrial time-and-motion breakthrough in his Michigan automobile factory in 1913. The addition of robots (by GM in 1962) alongside human workers has further increased efficiencies. But the success of the assembly line is tied to the quality of the components fed to it from production.

Thus, as the complexity of modern vehicles has risen dramatically in recent decades, particularly in the case of electric ones (EVs), the value of pre-assembly quality checks at key stages during active production has become increasingly apparent. In addition to various methods of surface scanning, some of the leading automotive manufacturers are using computed tomography (CT) to inspect deep inside complex components to ensure they are fit to send to the assembly line.

The Importance of CT Scanning

CT scanning is expensive. But so is only finding out that costly components are defective after a vehicle rolls off the assembly line. The savings achieved in time and resources when valuable automotive parts are inspected for quality before they are installed in a vehicle can be significant. That’s what German luxury automaker BMW has concluded – which is why they are CT scanning the engine housings of every single one of their advanced IX model EVs.

In contrast to the company’s other current electric vehicles, the IX is based on an independent platform. But it uses the same drive unit (motor, inverter, transmission) known as “Heat” (Highly-Integrated Electric Drive Train) that is employed in the other BMW models. Ensuring the consistency and quality of this valuable drivetrain is clearly a priority for the automaker.

Robots, Scanners, and Software – On the Line

At its Landshut, Germany, facility, BMW is using a combination of in-line robotics, CT scanning and detection, and analysis and visualization software, to inspect their “Heat” e-engine housing during active production. The entire system was installed and integrated into their production line by Heitec, a global engineering solutions provider for a broad range of industries. Heitec’s installation delivers a continuous-flow process in which a robot plucks an aluminum housing off the production line, deposits it on a platform that feeds it into a Heidetect Inline CT scanner, then shifts to the other side of the scanner to retrieve an already-scanned component and move that one back to the line. Each scan takes place in 50 seconds.

Post-scan, Heitec’s proprietary software quickly generates a digital model of each housing from the CT data, then transfers the model to Hexagon’s VG Inline software for rapid segmentation and analysis. Viewable onscreen by engineers, graphic-and-colored alerts are automatically generated to call out any significant porosity, cracks, geometric inconsistencies, or other production or material flaws that deviate from BMW’s specific metrics requirements – and might impact quality.

“Doing all this in the cycle time of the production line is the challenge,” says Heitec co-founder Christian Abt. “You have to speed everything up to stay within that time – or even several seconds less than that.”

Interpret Low Resolution Scans with AI

The solution to the time crunch? Employing lower-resolution industrial CT scans that need significantly less time to capture an image – plus Heitec’s scanner-integrated software designed to process this “less-heavy” information and reconstruct the component’s volume data into a 3D computer model. This data is then loaded into holistic analysis and visualization software that queries the digital model, employing algorithms and, most recently, deep learning – to reliably identify, interpret and report potential quality issues despite the noisy images – within a much shorter time window.

“For the kind of advanced data processing software that would deliver the metrology results our customers were asking for, we found our way to Hexagon’s VG Studio MAX and VG Inline,” says Abt. “The complete infrastructure of VG Inline is a ready-to-use software package that allows us to easily integrate our own modules and functions, enabling us to combine accepted standards with flexibility. We just drop out a volume and VG starts to analyze it. This process is also easy to parallelize to more than one computer to speed things up in case the time for analyzing the data is slightly longer than the time to scan the part.”

The VG CT-data analysis and visualization software has undergone over a quarter century of evolution and development, with process automation and, most recently, deep- and machine-learning, increasingly being integrated into its offerings. Says Hexagon’s general manager for VG products, Dr. Daniela Handl, “The huge quantities of design and engineering data that can be captured by CT need a holistic approach, coupled with robust computational capabilities, to be most valuable. Rapid results delivery is critical for engineers looking for real-time insights into production-line quality as well as how manufacturing-parameter variations might be affecting outcomes.”

Key to Segmentation and AI Integration

Key to determining which data is significant in the scan of a component is the process of segmentation – the extraction of regions of interest (ROIs) from the 3D image data. Abrupt discontinuities in voxel grey-values typically indicate edges that define a region. By partitioning a scan into discrete regions, the software can more rapidly focus on and catalogue the characteristics of the voxels in each segment. Segmentation thus enables the processing of only the important portions of a dataset – which takes a lot less time.

Customizing Proprietary Solutions with Machine Learning

But how do you determine which segments are the important ones? Especially since, in a lower-resolution CT scan obtained on a production line, the less-crisp image slices can be more challenging to understand. This is where deep- and machine-learning, forms of artificial intelligence (AI), comes into the picture: using trainable algorithms, AI can be harnessed to process and make decisions based on noisy CT data as accurately as on hi-resolution data.

“Trainable” is the key word here. The algorithms compare what they see against an existing database of all known identifiable defects that have been customized to each manufacturer’s product geometries (which is where the training comes in). The algorithms learn to identify specific manufacturing flaws by interpreting what the less-crisp voxels show.

Over time, deep learning gets even better at what it does; by comparing what it sees against real-world product data for which the solution is known, it learns to recognize patterns and features and call out deviations from the norm. In this way, it can provide highly accurate snapshots of what’s actively happening on a particular production line – supporting confident decision-making about whether to accept or reject a part. This in turn informs production-variable changes, the effects of which can be captured, collated, and statistically examined. EV batteries are another area in which machine learning can be used in this way.

How to optimize this kind of AI methodology is what BMW is now exploring. Working closely under NDAs, with innovators such as Heitec and Hexagon, these companies are creating and curating their own internal data sets with which to train in-house deep learning systems. “Training a deep learning system to one’s proprietary data understandably takes some time,” says Handl. “But it pays off by saving highly valuable time and resources on the production line.” Actionable insights into what’s happening on one’s own factory floor are contributing to the further evolution of smart manufacturing, enabling companies across many industries to identify ways to improve quality while making their products more competitive.