Machine-vision system detects engine-fuel leakage

Machine-vision system detects engine-fuel leakage

-Andrew Wilson

In its latest A Series automobile engines, Daimler-Benz (Stuttgart, Germany) uses a novel common-rail platform as part of the fuel-injection-system design. Instead of injecting fuel into each cylinder independently, the pressurized single or common-rail platform delivers fuel evenly to all four cylinders. Mounted at opposite ends of the common rail, a high-pressure pump and a pressure-control regulator direct the operation of the fuel-injection system.

To test the A Series engines, engineers Claus Lorcher, Hardy Burkle, and Hans Ramsperger at Daimler-Benz have developed a machine-vision system based on off-the-shelf components. Designed to check whether the fuel-injection system has been mounted onto the engine correctly, the system uses five color cameras that inspect the pressure-control regulator, high-pressure pump, fuel-supply connectors, and all four engine cylinders.

Before performing engine testing, a fluorescent dye is mixed with the fuel. Using an electric motor coupled to its drive shaft, the engine is then slowly brought to a speed of 1000 rpm. Illuminating the engine with a near-fluorescent lighting system from AEG (Hamburg, Germany) detects any fuel leakage as a green fluorescent emission.

To automate this inspection process, images from all five cameras are multiplexed into a PCI-based, color frame grabber from Data Translation (Marlboro, MA). Then using Neurocheck software from DS GmbH (Remseck, Germany), the histogram of the green fluorescent region of interest (ROI) is matched with a histogram of an image of known fluorescent emission. If matching occurs, a threshold of the ROI of the original image is used to determine the amount of emission. Here, pixel counting is used to determine whether to pass or fail the engine.

"Because there is always a tiny leakage in every engine," explains Lorcher, "we have to define a limit which is acceptable." If the system counts fewer than 20 pixels in each image, the engine is passed. If, for any reason, an engine fails the inspection, each ROI is logged automatically, and the operator can immediately find where the leakage occurred.

At Daimler-Benz, the machine-vision system is used in an on-line process that requires approximately three minutes to perform each inspection. Interestingly, it takes approximately two minutes to load each engine onto the inspection station and bring the engine to speed. To automatically inspect the engine, says Lorcher, takes about 30 seconds.