3D inspection systems visualize surfaces down to the smallest detail. This is necessary in the automotive industry, for example, for checking paint defects. Because even the smallest defects or inclusions in the paint lead to costly reworking. Deflectometers help to detect errors quickly and reliably and to process these automatically.

More and more industrial applications require reflective surfaces to be recognized and measured. For example, in electronics manufacturing in smartphone production or in the automotive sector when producing mirrors or checking paint defects on a finished car body. In order to detect a fault or defect on these surfaces, OEMs require systems that measure with high precision and detect deviations in the surface structure in the sub-micrometer range. Micro-Epsilon has developed the reflectcontrol sensor series, which is based on the measuring principle of deflectometry, for all these fields of application. Here, the sensor displays a sinusoidal striped pattern via a monitor, which is reflected into the sensor's cameras. The surface of an object being measured is, for example, a car body. The resulting phase images can be used to calculate so-called kernel images based on the measured variables of base intensity, amplitude and curvature.

High-resolution Measuring System

Micro-Epsilon’s reflectcontrol automotive is known as PSS 8005-D and has a measurement area of 367.5 mm x 823.4 mm. The sensors measure even the smallest defects with a resolution of 185 µm and work with a measurement data acquisition time of 400 ms. The supply voltage is 24 V DC with a power consumption of less than 200 W. The high-precision resolution benefits, for example, the paint defect inspection of car manufacturers. In the past, they relied on light tunnels in which skilled workers would use fixed lighting systems and cameras to detect defects in the paintwork. Alternatively, systems with static sensors, which the car body moved past on a conveyor belt to detect defects. However, both are options are very error-prone.

Suitable for Short Cycle Times

With the reflectcontrol system, the body is brought into a fixed position. The sensors are attached to robotic arms and move across the bodywork. Micro-Epsilon opted for this method because it achieves the lowest error rate and the required cycle times can be met. Two to four robot arms, each with a sensor, work at each inspection station–allowing the entire body to be measured. In order to accelerate and decelerate the robot arm sufficiently, it was necessary to keep the total weight of a sensor below 50 kg, which Micro-Epsilon achieved using a carbon housing. This also increases the robustness of the overall system and protects the robot arm. With the light tunnels described above, car manufacturers achieved a defect coverage of around 60 percent in the past. With reflectcontrol, fault coverage of almost 100 percent is possible.

Automatic Processing of Defects

First, the vehicle is measured from one measuring position to the next and any defects found are projected back onto the vehicle surface. A defect can be localized at +/- 3 mm on the vehicle surface. 3D features such as height, depth and volume are added to each defect using 3D reconstruction. All collected data is saved in an XML file and is thus available to the vehicle manufacturer. Once the defect has been detected and localized, automatic processing can begin, which Micro-Epsilon implements together with its partner companies Asis and Virtek Vision. This is done using additional robots with active force control, each equipped with a dual-mounted sanding and polishing head. A robot first grinds off the defect and then polishes it. Used sanding sheets and polishing sponges are removed at changing stations and new ones are picked up. Using a laser marking system from Virtek Vision, the defect is then marked as machined without contact by projecting light onto the body. If required, additional defects can be displayed via the projection, which must be reworked manually. The system knows at all times which robot is currently processing which points. The most important prerequisite for automatic processing is the database that Micro-Epsilon can provide with reflectcontrol. This enables OEMs to optimize their production, identify error chains and intervene at an early stage if errors occur.

System Learns Automatically

The detailed database also benefits the classification of 3D data. Defect detection results in a 3D reconstruction–in principle, it is based on the integration of the curvature data. It is sufficient to reconstruct in the vicinity of the defect. The system can independently define new defect classes. Due to the large database, the system can be used for audits. The labeled data trains also the underlying AI algorithms. All data is also available to the OEM in a file and can be read into the standard higher-level Q systems. This means that the data can be visualized using established tools. The system learns from all the data it collects and therefore becomes increasingly intelligent. This can also be used to set up early warning systems to detect and eliminate error chains at an early stage. For example, warning thresholds can be created when certain numbers of defects are exceeded. In addition, so-called heat maps can be used to find error hotspots.

Evaluating the Surface Structure

In addition to automatic processing, Micro-Epsilon can perform an appearance evaluation. For a structural evaluation of the surface, the existing paint structure is broken down into its spectral components. The components are summarized in different frequency ranges, for example, which short and long waves are contained in the structure of the surface, because no surface is perfectly smooth. It is important for car manufacturers to maintain certain appearance values on a car body. Automated cells, end-of-line cells or aqua cells, are normally used for this purpose. A vehicle is inserted into this and viewed by a robot with a scanner. The robot moves to the various positions and measures the appearance directly over the surface. This takes about 20 minutes in total - so a maximum of three vehicles can be measured per hour. With the system from Micro-Epsilon, the appearance can be evaluated on the basis of the existing images without additional hardware. This saves time and costs for the OEM.

Extensive Database

Another important innovation is the classification of data. This can be used to automate the processing of errors. Classification means recognizing the type of defect, for example, whether it is a crater, an inclusion or a defect from the subsurface. The customer also has access to 3D data–so they can access the exact dimensions of the defect at any time and initiate extensive analyses. As all data can be transferred to the company's internal Q-System, alarm functions can be set up and faults communicated at the appropriate point.

Conclusion

With Micro-Epsilon reflectcontrol, defects in reflective surfaces can be detected and classified almost 100 percent. The system is superior to conventional light tunnels and static systems, particularly in the field of paint defect inspection in automotive production. Due to the large database, the manufacturers of automatic processing systems are able to assign the correct processing recipes to the defects. This reduces unnecessary defect handling and thus saves resources and costs.

Author
Konrad Steinhuber, Group Management Project Planning Surface Testing at Micro-Epsilon