Magnetic crawler for welds Visual Testing, based on 3D profilometry and 2D image processing

Abstract

To ensure the satisfactory performance of a welded structure, the quality of the welds must be determined by adequate testing procedures. During Visual Testing (VT), the weld is examined through the eyes of an inspector to determine surface discontinuities. Although this is commonly considered the easiest, quickest, and least expensive type of inspection, it has some limitations: a certified inspector shall always be on site to perform the test, making it strongly dependant on his experience, knowledge and current environmental conditions. In addition, there is no possibility to record inspection data except standard pictures and comments to be added to inspection reports. Modern technology brings novel solutions for quality assurance. This article describes a powerful tool which can improve the reliability of VT that combines precise laser measurements, image processing and data cloud computing. A portable magnetic crawler has been developed using a Raspberry Pi SBC, a smart profile sensor and a 5 MPix industrial colour camera in order to gather both weld 3D point cloud and surface pictures. Laser triangulation and processing power directly integrated on board allow an easy weld profile measurement. Undercuts, reinforcement excess, spatters, hi-low are precisely detected and sized by the laser sensor while integrated 2D camera records and analyses surface features such contaminations, corrosion, and weld discoloration. The point cloud weld reconstruction is realized stitching together 2D profile data at a fixed interval that can range up to a hundredth of a millimetre with a spatial resolution up to 50um. Inspected weld is fully digitalized in the form of 3D object together with a set of numerical features, data are securely stored in a local flash memory and automatically synchronized with cloud platform for remote data visualization. Automatic weld assessment can also be applied using cloud computing and artificial intelligence (AI) algorithms together with deep learning tools. The combination of images, 3D models, measurements, and other evaluable features all together in the same digital platform definitely enriches the weld assessment making it quicker and reliable with the support of AI, it gives the possibility of remote visual inspection, it ensures a flawless workflow where visual inspection data are stored in a dedicated platform and can be accessed at any moment by stakeholders.