Spookfish has deployed a vision-based inspection system to detect anomalies in core housing brackets manufactured for turbochargers. In combination with a robot responsible for handling the component to be inspected, the vision system is able to inspect five faces of the component within a processing time of less than 10ms per face. The system has been successfully trialled for four variants of turbocharger housing brackets.
The high success rate and increased development of turbochargers in the 21st century has resulted in compact, complex and efficient assemblies that meet increasing technical demands. On the other hand, to meet market demands, the manufacturers need to consistently ensure excellence in the manufactured components, while also maintaining a high level of efficiency in the manufacturing process. In order to fulfil these demands, Brakes India have completed trials of a high speed vision-based inspection system developed by Spookfish, leveraging bespoke computer vision algorithms to automatically examine the housings and detect defective parts.
The initial critical measurement carried out by the software is the angle of rotation of the base of the component, which is transmitted to the robot within 3ms after the component is presented to the camera. The transmitted angle is then used by the robot to precisely position the other four faces of the component for inspection by the vision system. Apart from the angle measurement, the presence of a half ring is verified, and a diameter check is carried out. Inspection results from one such base, from one of the four variants, is shown below.
Once inspection of the base is complete, the robot presents each of the other four faces to the vision system, which then verifies the correctness of several parts in each face. For instance, in the first face the system verifies the presence of a pin towards the top of the assembly, and also checks that two holes have been threaded and are ready for the next stage of assembly. Thanks to the use of efficient (while also accurate) algorithms developed by Spookfish, the inspection process for each of the faces is completed within 10ms. If all parts are verified for correctness on all faces, the software sends a ‘Pass’ signal to the robot, or a ‘Fail’ signal otherwise, along with a visualisation of the correct and defective parts.
So far, Brakes India have used Spookfish’s vision system for inspecting four different variants of turbocharger housings. The images displayed below show some examples of the process across the four faces.
The customised image processing software caters to a diverse range of customer requirements. In this instance, the software was heavily customised, with the pass/fail result for the entire component depending on a complex set of rules. Among the 4 variants catered to, each consists of 5 faces with different sets of inspection requirements.
Spookfish prides itself on providing high quality solutions to challenging problems, and this project was no exception. The system compensates for small shifts in translation and rotation errors which are inevitably introduced through the automation process. Most machine parts are inconsistent in terms of appearance, and can often pick up significant amounts of dust, grease and surface irregularities. There are also considerable variations in lighting conditions. However, Spookfish’s vision team excels in analysing and addressing these issues, customising their algorithms to tackle several scenarios.