Pressing Shop Floor Questions: Challenges & Solutions
The all about automation Friedrichshafen was once again a great success! Crowded aisles, engaging discussions, and an energy that truly inspired me. It's always refreshing to speak directly with you – the practitioners – and hear the most pressing questions and challenges from the shop floor.
From countless conversations, the team identified the most common questions and concerns currently on your minds:
"We sometimes develop solutions ourselves but fail to implement them. And digitalization and networking are missing!"
- Our answer: This is a classic. Many special machine builders or production companies start with in-house developments but quickly reach their limits when it comes to scalability, robustness, or integration. The key lies in a robust architecture and a unified data basis. STIWA offers proven platforms that close these gaps and enable true networking – without you having to reinvent everything from scratch.
- However, if you still want to continue building yourselves, we offer a stable foundation for your own developments.
"How can we relieve our employees directly on the line and prevent errors?"
- Our answer: Manual Data Input, start with transparency, even if you largely produce manually.
- Many are looking for solutions to react flexibly in daily production, but based on data and not emotions and stories like "but I heard...".
- Our system impressed many precisely because it delivers just that!
"Is our own OEE system still future-proof, or does it require external expertise for the next leap in efficiency?"
- Our answer: Some of you already have your own OEE systems. But many are considering whether external solutions would bring more benefits. This is about measurability and business case. Systems like our MONITOR and OPTIMIZE not only provide transparency but also deliver a clear business case. They are designed to give you the biggest leverage for efficiency improvement.
“How do we connect legacy systems (brownfield) with a modern control system and use AI for analysis?”
- Our answer: This is the supreme discipline, as experienced experts confirmed. The questions revolved around connectivity and interfaces to machines and controls. From a control system for machine tools to AI-supported analysis. Our software is designed to meet these complex requirements and continuously use your data from the machine to management – for transparency, analysis, and forward-looking decisions.
A central point, also reflected in your active participation, was the question of the optimal time to start data collection in production. When Philip asked whether usable data should be collected before or after commissioning, 90% raised their hand for "before commissioning." This is a clear and very positive sign that awareness of the immense importance of an early data strategy is already widespread. Many have obviously understood that "data crude oil" cannot be refined only later, but that its extraction must begin in the early phases of project planning and engineering to become usable "fuel" at all.
The answers to Philips survey provide further insights into the reality of many companies:
PoC Frequency vs. Rollout Success: On average, participants implemented 2.8 PoCs last year. However, the "brutal truth" emerged when looking at the rollout: only 4.8% of these PoCs were successfully rolled out on a large scale. This figure emphatically underlines the gap discussed in the lecture between a successful Proof of Concept, which takes place under ideal conditions, and the complex reality of a global rollout, which requires comprehensive infrastructure, cultural, and data strategy adjustments. A PoC is often manageable, but a rollout is a mammoth task.
Understanding AI and LLM Concepts: The survey paints a clear picture regarding the understanding of modern technology concepts:
Terms like "context window" (average value: 3.5 out of 5) and "system prompt" (average value: 3.6 out of 5) are already somewhat familiar.
"Semantics" (average value: 3.9 out of 5) indicates a tendentially higher level of familiarity.
In contrast, "RAG" (Retrieval-Augmented Generation) with an average value of 2.0 still appears to be a largely new field for most, which could indicate a future need for information and training in this area.
Many thanks to everyone who visited our booth and took the time for these valuable insights! We look forward to further discussions.
