SICK Automation recently supplied a local system integration specialist with an intelligent solution for its product inspection application. The solution – the dStudio web service and the Deep Learning software from the SICK portfolio – has proved highly accurate and reliable, creating a more accurate, efficient and productive inspection system.

This is the first AI-powered application for SICK South Africa. The customer, Jendamark Automation, was building a ring gear and diff cap assembly inspection system for its client, using SICK InspectorP621 camera sensors. The system was required to identify ring gear with missing bolts, and incorrectly assembled diff caps. But it had limited success in training the InspectorP621 system to identify bolts and diff caps correctly (a fruitless process that took several weeks). The highly reflective nature of the products and the low lighting of the plant environment meant that the InspectorP261 sensors failed to function effectively, as they could not “see” the products.

SICK Automation market product manager and market applications engineer, Anton Bresler, consulted with the customer over several days to provide technical support during the programming process. Realising the shortcomings of the existing system, he proposed the utilisation of SICK’s dStudio together with the Deep Learning software.

dStudio is a web service which enables the user to upload pre-sorted product images to the cloud. The Deep Learning software has the capability to analyse the images and “train” the user’s neural network (comprising SICK sensor systems) to make decisions, quickly and accurately.

Bresler worked alongside the customer, providing training on the functionality of the dStudio web service. He also guided the customer’s selection of product images for their system’s neural network – a process that took less than two hours. Following the successful implementation of the AI-powered neural network, programmed by the Deep Learning software, the customer’s inspection system was able to successfully identify individual product assemblies, even in conditions with low lighting. This has allowed the customer to automate the system’s inspection process, achieving greater productivity and efficiency, without the need for technicians or engineers to supervise or manage the process. “dStudio and Deep Learning deliver an inspection solution with very high reliability, delivering faster and more accurate processing. The fail rate for Jendamark Automation’s inspection system is now almost zero,” explains Bresler. This has proved a cost-effective solution, which delivers measurable ROI, fast.

dStudio and Deep Learning have been designed for use with SICK sensors. Both products are intuitive; users do not need technical skills or AI knowledge to complete their implementation. However, the SICK team provides user training and technical support to enable seamless adoption of the technology.

“dStudio and Deep Learning are ideal for a wide range of industries. In the automotive and FMCG industries in particular, where products are highly reflective and accurate inspection is critical for ensuring product quality, the solution is especially well-suited,” concludes Bresler.