Eliminating Automotive Defects Per Million With Automated Visual Inspection

Thirty-four United States presidents are equipped with cameras and ready to prepare high-performance shock absorbers and suspension systems for installation onto vehicles in the performance and luxury auto market.

That might sound strange, but at Thyssenkrupp Bilstein’s Hamilton, Ohio, facility, the collaborative robots (cobots) that help with automation projects and part inspection are named after U.S. presidents. The manufacturer of shock absorbers and suspension systems made its first attempt at automation in 2016. Thirty-four cobots later, Bilstein has begun using its cobots to automate inspection processes throughout the facility to avoid human error during manual inspection.

Cam-Pains With Cobots

Bilstein shared one primary pain point that led to this change during Modern Machine Shop’s visit to its facility in May 2023. The facility assembles a lot of very small components necessary for installing its shock absorbers onto a vehicle, and too many of these components were missing.

A key performance indicator for the automotive industry is the parts per million (PPM), a measure of the number of defective parts per million produced. When parts were inspected manually during eight-hour shifts, missing components on parts that left the plant accounted for the company’s highest PPM issue. 

“One incident creates a situation where you have to put containment on everything,” says Bobby Adams, quality director at Bilstein. “So they’ll go to a tier one (vendor) typically to get sequenced, get a VIN number attached to it, and all of a sudden you’ve got a damper that falls out and it stops the process. So that’s where the pain point really comes in.”

With this issue, Bilstein’s Industry 5.0 team, which includes controls engineer Craig Mathews and manufacturing engineer Herman Baker, was tasked with creating an inspection solution at the end of the assembly line that would eliminate PPM.

Lights, Camera, Inspection!

Mathews and Baker came up with the idea to equip Bilstein’s cobots with Keyence and Cognex vision systems to automate the inspection process after completing part assembly. Each part has a label that is scanned by an operator so the inspection system knows what to look for on each set of parts. The visual system identifies specific combinations of pixels to determine if components are present. If required components aren’t present, the intended pixel markers will not be in the image, triggering a failed inspection.

The vision systems on each cobot scan labels on completed parts to ensure assembly is fully and properly completed. This process includes several checks to determine if the wire harness is properly installed, the spring seats that damp the energy from bumps are present, and any covers and caps are fully pressed onto the part.

On the first iteration of this solution, Mathews and Baker realized there may be potential issues with lighting. Throughout the day, ambient lighting angles are changing along with inspection angles as the cobot shifts positioning to inspect different areas of a part.

“One of the problems with having the camera moving is you have five-to-10 inspections and have five-to-10 different possibilities of the light being different,” Mathews says.

With the second-generation setup, the camera now takes multiple different types of strobes. Instead of flashing a light that may look different every time based on the measurement angle, the camera uses different sequences of lighting, and an operator can choose what the best look is. Additionally, these vision systems offer filters that can reduce ambient lighting influences on the shots.

“Before this, we had a fixture plate that was designed for each fixture, and when we had a new variant come along, we’d create a whole new fixture plate,” says Brian Williams, COO and director of operations at Bilstein. “We wanted to replace that with something new, which was the first generation. The second generation enabled us to have infinite programs and infinite fixture plates changed by just typing in a new part number, and the robot or the computer knows everything it needs for that particular model.”

The automation of this process is touted by the Bilstein team for its consistency and repeatability. The cobots can move around to measure multiple positions and hit those positions with consistency every time. Before, an operator might have had to shift the part around to several different inspection positions or inspection stands to properly inspect it. A robot equipped with a camera will never get fatigued from performing the same task, and it will do every check the exact same way every time, eliminating most of the issues with missing parts or small pieces falling out without getting noticed by an operator.

To test the effectiveness of these processes, Bilstein operators intentionally built shock absorbers incorrectly and had them inspected by the cameras to make sure they fail the inspection checks. When inspections are failed on the floor, the parts are placed in a separate box away from the parts that pass inspections.

Data Tracking for Accountability and Traceability

To keep track of its inspection data, Bilstein uses data matrix codes to encode data and track when parts were inspected, where issues occurred and when parts were shipped out. This information was previously stored in boxes, but the company recently invested in better data tracking and traceability. Bilstein now keeps a year’s worth of inspection data on hand throughout the facility. This allows operators to compare performance and inspection metrics to see where improvements can be made.

Older data is archived and stored on a cloud server. With each aspect of the inspection tracked and stored, Bilstein can quickly determine where — and if — any issues may have occurred prior to parts getting sent to a customer if the facility receives a complaint.

“Before we start to get into containment or before we start to panic, I can immediately tell [the customer] if anything happened and what happened, or if there is a non-issue,” Adams says. “That’s the kind of power you get, and you build in that efficiency and timeliness.”

Prior to tracking and storing this inspection data, Williams says Bilstein would typically have to accept responsibility for missing parts or other issues that led to complaints. With 100% traceability of its processes, the company has the data to work collaboratively with customers to identify the root cause.

“It challenges us to be better, and it also challenges them if maybe they’ve got a gap in their process that maybe they can also tidy up,” Williams says.

Pairing this automated inspection system with traceable data has paid dividends for Bilstein, nearly eliminating the facility’s biggest pain point: the PPM complaints from customers. Adams says inspection quality with its Keyence vision systems is nearly 100% accurate now, whereas visual inspections by an operator were around 80% accurate at best.

“The old issues that we had that produced PPM complaints from our customers have reduced to maybe one every quarter, maybe one every six months, whereas we were previously probably hitting two or three per month,” Adams says. “It’s just simple things like the orientation of a plate, missing a component or a component falling off in transit or mishandling by a customer. Those things happen. It’s part of the process, but being able to identify it and resolve it, that was the key for us.”