The AIoT Lab Heilbronn is operated by Ferdinand-Steinbeis-Institute at the Research Campus in Heilbronn. This setup ensures access to the academic network in Heilbronn, which can be easily extended to include the greater Stuttgart area and Baden Würtemberg.
Motivation and Goals
The Lab developed at the Bildungscampus in Heilbronn at the beginning of 2021. The primary motivation is to provide smart connected products and solutions that illustrate how the challenges in real world can be overcome. Technological as well as practical expertise is offered in the duality of research and entrepreneurship in order to create an environment that reflects the aforementioned.
The solutions developed in the AIoT Lab closely follows the AIoT PlayBook. Experts around the world have come together in order to develop the concepts in a more understanding way that helps us to build a smart connected product and solution to the challenges that is present in the industry.
The AIoT Lab has built an initial demo system which utilizes AI-based ultra sound analysis to detect leakages in pneumatic systems. The goal is to use this platform as a real-world demo to evaluate concepts in the following areas:
- Holistic DevOps: Validation of best practices for an integrated DevOps model which combines cloud, edge and AI
- AI/ML model re-use: how can re-use of existing models be maximized, especially from the point of view of smaller companies with limited R&D capabilities.
By Using the AIoT Lab, we determine to
- Get early validation of feasibility.
- Reduce development costs.
- Re-use of AI models for IoT.
- Ensure efficient AIoT infrastructure.
- Benefit from collaboration with AIoT User Group.
Typical collaboration models is to support
- FSTI micro testbeds(validation/feasibility)
- Funded research projects (solution design/proof-of-concept)
- Support transfer of knowledge from academia to the industry.
Our Use Case
Pneumatic Systems is the use case considered in our AIoT Lab. A pneumatic system is a system that uses compressed air to do work for automated equipment through a combination of inter-connected components. Examples can be found in industrial manufacturing, car wash, food industry or medical applications. Linear or rotary motion are the forms in which work is usually produced. To protect the cylinders, actuators, pipes and pumps, filtering and drying of the compressed air or the pressurized gas are usually implemented.
20-30% of a compressor’s output can be leaked which can account for a significant source of wasted energy in an industrial compressed air system. A leak rate equal to 20% of the total compressed air production capacity is the usual expected loss due to leaks in a typical plant that has poor maintenance. According to the company Madder, around 62,000 compressed air systems are installed which consumes 16 million kWh Energy approximately. Hence detecting such leaks will lead to maximum saving potential around 50 % in this area. Leaks are also one of the primary contributors to operating losses other than being a source of wasted energy. A drop-in pressure that is caused by the leaks can make the system less efficient, which in turn adversely affects production. Common leak points are: Couplings, hoses, tubes, and fittings, pressure regulators, open condensate traps and shut-off valves, and Pipe joints, disconnects, and thread. In our Lab environment, we use the LEGO- pneumatic system to develop the prototype.
Our AIoT Stack
The Dataflow within architecture is cyclic and the Edge Device is the key to the Data Flow.
Edge Devices are the end nodes typically located on the observation site. As it is a crucial part of the IoT architecture, the computation is supposed to be ubiquitous in nature. For the current Pneumatic use case, the design decisions are Harmonized with the most Successful IoT device design. I.e. A palm sized; touch screen based computing device. User can interact with the UI designed on the edge device. (In case of invisible Ubiquitous Design Pattern, the user interaction will be different) In the Pneumatic use case, we are using a touch screen to render Kivy based UI.Edge devices also need to perform the necessary signal processing i.e., signal acquisition, signal conversion, denoising, feature extraction. Edge device should also be capable of running a fully trained neural network.
Communication is often inferenced as interpretation. For the proposed architecture, the interpretation is part of the Edge Device's user interaction. And Communication is the way device interacts with the system.
This part of the architecture handles the internal essential operations such as Database Management, User Handling, File Storage, Device Monitoring as well as the service offerings such as Visualization.
- DB and User Management – To manage and Modify UserData and necessary Metadata
- File Storage – Storage of System specific files such as weights, data, datasets, models etc.
- Device Monitoring – To keep on the monitor processes for the deployed edge devices
- Visualization – Providing graphical insights to the data as well as device alerts
Our AIoT Lab uses the tool - LeanIX that offers SaaS solutions to help IT architects, IT asset managers, business leaders, and DevOps teams achieve transparency and control over their enterprise architecture, SaaS, and microservices landscapes. The tool becomes important in maintaining the value stream of the project.
In cooperation with the company Mader GmbH and StackIT, a first prototype launched by our team at AIoT Lab. The requirements of robustness, reuse and fairness are in the foreground. These were defined together with Mader GmbH & Co. KG. Pneumatic systems are used in countless applications today. Any leak in a pneumatic system leads to increased operating costs and energy. According to the Mader company, Around 16 million kWh of energy is wasted per year and detecting such leaks as a saving potential around 50%. The solution approach piloted in the AIoT Lab thus contributes to significantly increasing both profitability and sustainability.