Part 1: OEE Capture & Manufacturing Intelligence
oee.ai
basic philosophy
oee.ai sees itself as a manufacturing intelligence start-up founded in Aachen in 2016. The core conviction: AI and algorithms support employees — they do not replace them. Domain knowledge among the workforce remains essential, because even if algorithms suggest improvements, a person must validate and implement them.
OEE in an FMCG context
In the fast-moving consumer goods sector, two of the three OEE factors dominate the loss structure:
Availability level: Set-up processes, technical losses, planned and unplanned shutdowns.
Performance level: Often underrated. For example, anyone who could fill 120 pieces/minute but only produce 60 has an efficiency of 50%. From practice: Performance losses are often greater than expected and are not systematically recorded by many companies at all.
Quality level: In the FMCG sector, typically at 98—99% — important, but rarely the biggest lever.
Three product components
NanoMes: Lightweight manufacturing execution system, plug & play, software as a service, no IT project, no IT resources on the customer side. Installation in one hour.
Reporting & visualization: Widget-based live reports, individually configurable — from technical shift reports to factory reports for production managers. Andon board, industrial smart watch, shop floor integration
Advanced analytics: Algorithms for loss forecasts, anomaly detection, stability figures and AI-based action suggestions.
Data collection: Three ways
Direct connection: Modern systems with OPC UA or MQTT — low effort, no additional hardware.
oee.ai sensor box (plug & play): Proprietary hardware with upstream SICK sensor for quantity recording, transmission via LTE to the data center in Frankfurt (European data protection law). Over 80% of customers use mobile communications instead of WLAN.
Connectivity partners: Can be retrofitted for all types of PLC control systems.
The central data element is the quantity vector — pieces over time. Everything else (availability, performance, OEE) is calculated from this. Non-discrete units such as meters or square meters can also be imaged.
Loss reasons that the system cannot know itself are recorded via a tablet with an adaptable multi-level loss base catalog — separated for availability, performance and quality losses. The unique selling point: a full range of organizational and procedural losses, not just technical reasons.
Visualization & evaluation
- Heartbeat line: Target vs. actual output in real time, downtimes in red, power losses in orange
- Shift comparison: Early vs. late vs. night shift
- Factory overview: Several investments aggregated in a waterfall diagram — where do the biggest losses add up?
- KPI tables: Exportable via CSV, API access for Power BI or other systems
- Andon Stories: Context-sensitive board sequences — e.g. during an ongoing set-up process: Algorithm shows the trend of the last ten set-up processes and gives the employee a time window target
AI use cases
Prognosis: After around six months of learning, the algorithm recognizes patterns of losses and predicts the probability that the investment will be standing again in X minutes. Basis for the maintenance technician to decide: stay at the plant or go?
Scrap detection: Algorithm identifies break points in the scrap curve (e.g. depending on batches of raw materials), automatically triggers a push notification and suggests which measure helped in the last similar case. In the example shown: approx. 7,000 meters of waste could have been saved.
Getting started: Proof of Value
Two months, two systems, two sensors, two tablets. Preparation: Coordination on shift systems and products. Setup & training: approx. one hour until the first data point. Conclusion: Analysis workshop — what was found, which improvements can be initiated?
Part 2: Digital shop floor management
SFM Systems
What is digital shop floor management?
Not the 1:1 digitization of the whiteboard — but a new way of thinking: providing information tailored to target groups and applications, leading meetings in a structured way, methodically supporting problem solving. Objective: Support managers and employees in making decisions based on data.
SFM Systems, founded in 2018 from TU Darmstadt, combines lean management methodology with digital infrastructure.
The core features
Performance dialog: Structured shop floor meeting with configurable agenda (attendance, key figure dashboard, list of measures, discussion topics). Key figures are automatically drawn from various sources — oee.ai, SAP/ERP, quality systems, manual entries.
Deviation management: Automatic recognition of discrepancies from all stored key figures. As soon as a target or limit value is exceeded, a digital workflow starts: describe variance → analyze → derive measure or initiate problem solution → escalate or complete.
Structured problem solving: Step-by-step guide through 8D, Ishikawa, 5×Why, or Closest Logical Comparison. Several people can work on it at the same time, directly on the tablet on site. All problem solutions remain documented and searchable — even six months later.
Process confirmations: Recurring checklists are automatically assigned to employees. Negative answers require a comment and can directly generate an action or discussion topic. Managers can see at a glance whether checklists have been carried out and what the result was.
Knowledge store: Every measure, every problem solution is permanently archived — with search and filter functions. Building up knowledge is an explicit system goal.
data integration
SFM Systems uses an industrial IoT platform that can connect SQL databases, REST APIs, OPC UA, MQTT, SharePoint and external applications. Available as cloud or on-premises. oee.ai is one of the supported data sources.
AI outlook
Current projects: Combination of process data (plant data) with action data (what was done to eliminate losses). Objective: Predict how key figures will develop and recommendations as to which measures have previously worked in similar cases.
From the Q&A session
Is it complex to connect additional systems? Depends on the available interfaces. The better data is already prepared, the easier it is to integrate. oee.ai and SFM Systems are actively working to simplify their coupling.
Unique selling point oee.ai? Simplicity of recording (plug & play, one hour until the first data point), complete loss spectrum (availability, performance, quality — organizational and technical) and the breadth of evaluation from widget reports to AI analysis.
High quantities (six digits per shift)? Not a hindrance. Recording via sensor or directly via PLC control — technically easy to implement.