Aside from a few pioneers, smart factories to date have been relegated to being a concept only. While a few companies were just beginning to consider Artificial Intelligence (AI) to take their production to the next level in terms of efficiency and flexibility, Covid-19 came along and pushed them out of their comfort zones. Even Gartner analysts expect smart factories to become mainstream within the next two to five years.
Two people who very quickly sensed the potential and seized the opportunity were professors Dr. Franz Haas and Dr. Rudolf Pichler from the Institute of Production Engineering at Graz University of Technology (TU Graz). They started a consortium of experienced manufacturing solutions providers as part of a call for proposals from the corresponding Federal Ministry (BMVIT previously, BKI now) to set up a real “smart” pilot factory.
According to the scientists, TU Graz had a choice of two options while setting up the Smart Factory: “Either to rely on a complete suite from a single major vendor, which would result in a vendor lock-in, making it difficult to use third-party solutions; or adopt a best-of-breed approach by using specialized partial solutions from different providers.”
TU Graz chose the latter option. The Smart Factory would have components provided by a consortium of service providers: enterprise resource planning (ERP), product lifecycle management (PLM) and manufacturing execution system (MES). The initial scope for T-Systems was to provide cloud expertise to support Big Data analytics along with a comprehensive Security solution comprising Security Information and Event Management (SIEM) and Security Operations Center (SOC) services.
However, the consortium soon came upon a major roadblock: how will the various solutions be integrated with the plants and machines so that the added values of the Smart Factory are realized? How to achieve a seamless production process through manufacturing, quality inspection and assembly to labelling, packaging and shipping – without any manual intervention?
The Smart Factory, which covers more than 300 square meters, started its operations at TU Graz in April 2021 and demonstrates the advantages of completely integrated manufacturing using the live example of a shaft gearbox for a robot arm. Prior to that, the project faced a rather tricky issue. The Smart Factory uses Computerized Numerical Control (CNC) machines that can automatically produce workpieces with high precision, even for complex shapes. But these high-precision machines need exact tools to match their processes and requirements. For this purpose, a tool presetter is used to allow the appropriate tools to be put together from a selection of different tool attachments. Employees need to select the attachments that match the respective program from a pool and mount them in the machine, making it a highly manual and error-prone process. A time-consuming comparison with the system identifier of the tool has always been necessary. An incorrect tool can not only spoil a complete series, but also damage the machine.
“The problem now is that, for the worker doing the assembly, the tools on the trolley cannot be clearly assigned to the program”, Michael Schmollngruber, Team Lead BI & Big Data Consulting at T-Systems Austria, explains. “There is a missing link from the tool to its identifier in the system at this point in the process”. This also brings the fifth V of Big-Data — Veracity — into play.
To solve this problem, TU Graz uses an AI-based object recognition service, which ensures that the right tool is installed for the program in question. To achieve this, employees currently still use a smartphone and take a photo of the tool via an app. This photo is sent to a cloud backend (Azure), where a specially trained algorithm identifies the tool shown in the photo and matches it with the NC programming. Within a few seconds, the employee receives a verification or falsification — whether they have selected the correct tool and receives further information via the ID of the tool, such as the upcoming production orders in which the tool is used or information from the PLM system.
The solution speeds up the selection of the correct tool, avoids rejects and quality defects, and significantly increases production efficiency. It can also be expanded to larger tool pools at any time or, for example, converted into an augmented reality variant that makes the use of smartphones unnecessary. Deutsche Telekom has already provided the necessary 5G campus network in the Smart Factory at the Graz University of Technology.
Container technology Docker is used in the backend. When the request reaches the Azure Cloud, the complete object analysis functionality is started serverless — and switched off again after identification is complete. In this way, the application is provided on demand. In addition, it can scale effortlessly — if multiple requests arrive at the same time — so that fast response times are achieved for all users. The AI algorithm used is an in-house development by T-Systems. In contrast to the usual approach, no real tools were used to train the model. The team took advantage of the fact that every object in the Smart Factory has a Digital Twin: the training data was generated from the CAD models of the tools.
The expertise of T-Systems around seamless data transfer, security, and data analytics in production makes the cooperation with the Telekom subsidiary particularly valuable for us.
Rudolf Pichler, Head of Institute of Production Engineering at Graz University of Technology (TU Graz)