Smart factory for FMCG manufacturers: Digital OEE and shop floor management improve performance

Part 1: OEE & Manufacturing Intelligence

Jörn Steinbeck, oee.ai

Starting point: Lean meets Industry 4.0

According to a BCG study, only 5% of companies combine a high level of lean maturity with a high level of Industry 4.0 maturity. It is precisely this area that provides the next level of operational excellence: Algorithms suggest productivity potential, real-time data accelerates decisions, sensors optimize processes such as set-up processes, and employees are specifically integrated rather than replaced.

OEE in an FMCG context

OEE = availability × performance × quality — all three as percentages. In the FMCG environment, the focus is typically on availability and performance, as the quality factor is often as low as 98—99%. Availability and performance, on the other hand, are often in the 50-60% range — this is where the greatest leverage potential lies.

Special situation: Performance factors above 100% (e.g. 114%) do not show above average, but incorrect default values from work preparation. The system automatically points this out — the first step of any OEE optimization is to clean up the master data.

The three product components of oee.ai

NanoMes: Lightweight manufacturing execution system, plug & play, cloud-based (server in Frankfurt, European data protection guidelines), no IT project required.

Reporting & visualization: Real-time cockpits, Andon board, industrial smartwatch, configurable widget interfaces.

Advanced analytics: Statistics and AI for forecasts, suggested actions and anomaly detection.

Data collection: Three ways

Direct connection: The system supports OPC UA or MQTT — low IT costs, no additional hardware.

IoT gateway: Proprietary hardware from oee.ai requires power and transmits via WLAN or mobile radio. More than 80% of gateway customers use mobile communications, as stable production WLAN is rare.

Hybrid: ERP/PPS delivers production orders and target speeds; reasons for faults that the plant cannot identify itself (staff shortages, missing orders) are manually recorded via tablets.

The central signal is the quantity vector — how many pieces were produced in which time increment. Availability and performance are calculated automatically from this.

AI applications

Predicting sequelae: The algorithm learns the typical system behavior before and after faults over approximately six months. He can then predict with a probability when the next fault will occur — and whether it is worthwhile for the maintenance technician to remain at or leave the system.

Scrap detection: The algorithm automatically identifies break points in the scrap curve (e.g. depending on roles used) and notifies the employee via push notification. In addition, it is proposed which measure worked in the last similar case — in the example shown, around 7,000 meters of waste could have been avoided.

Getting started: Proof of Value

Two months, two systems, two sensors, two tablets. With a simple connection, data integration takes half an hour. The biggest effort is onboarding employees and configuring reporting — a total of one day. After that, two months of testing with our own data.

Part 2: Digital shop floor management

Christian, SFM Systems

What is digital shop floor management?

More than just a digitized board: It's about data integration in the background and user-based preparation for employees and managers — for faster, data-driven decisions at every level of the shop floor cascade. The software runs hardware-independently in the browser (Chrome, Firefox, Edge) — on large shop floor screens, PCs or tablets.

The three core functions

Deviation management: Disruptions and deviations — both in systems and in organizational processes — are recorded, analyzed and transferred to a structured workflow. The system shows when a problem occurs for the umpteenth time and suggests starting a structured solution.

Performance dialog: Configurable shop floor meetings with automatically filled key figures from various sources (oee.ai, SAP/ERP, manual entries). The agenda, attendance tracking and key figure dialog can be freely compiled.

Structured problem solving: Guided process by 8D, Ishikawa, 5×Why — multiple people can work on it at the same time. Measures are given deadlines and responsible persons, can be escalated and are documented in the long term. In the case of recurring problems, it is possible to understand what was done the last time they occurred and whether the measure was effective.

Knowledge store: Every resolved fault, every measure is archived and can be set up. New standards from problem solutions are stored as process confirmations (checklists), which are automatically due at configurable intervals.

AI outlook

Ongoing projects with automotive manufacturers: forecasts based on historical KPIs (shift planning, order rescheduling) and recommendations from the action history — similar to a weather report for production.

Savings potential

The system saves time when preparing and following up shop floor meetings: Key figures are automatically filled in and reports are automatically aggregated. Leaders can focus on analysis and discussion instead of compiling data.

From the Q&A session

Predictive maintenance and OEE: Predictive maintenance addresses the unplanned loss of availability due to technical faults — a sub-element of OEE. oee.ai itself is not the deepest specialist here; specialized providers are recommended for complex predictive maintenance requirements.

Automatic fault detection without tablet: In principle, it is possible via an API connection to the system control system. In practice, a hybrid solution is being created: Automatically detectable faults come directly from management, organizational disruptions (staff shortages, missing orders) remain with the employee. Experience has shown that it is not possible to completely dispense with manual input.

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

VIDEO 13 · Digital Workshop Hoffmann Group

Webflow URL: /video/digital-workshop-Hoffmann-GroupSource: Presentation by Florian Langen, oee.ai

Lead-in suggestion:

Florian Langen shows how oee.ai brings machine digitization and manufacturing intelligence together in practice — from the first data point to the structured improvement process.

Transcript:

1. Initial situation in companies

oee.ai encounters three typical starting points in practice: companies that do not yet record any machine data; those that already record data but still work with pen, paper and Excel; and companies with existing MES landscapes that want to take the next step. The most common case is one of the first two.

2. The goal: Make productivity visible in real time

oee.ai has a clear goal: Every plant should be visible in real time — and if there are discrepancies, the right people are automatically informed (shift manager, production manager).

The target image can be built up in concrete stages:

Automated machine data collection — Is the machine running? At what speed? Target/actual comparison in real time.

Interference detection — When a machine is stationary, why is it stationary? The employee's domain knowledge is incorporated into the fault index via tablet entries.

Transparency for the shop floor — An Andon board (large television in production) gives employees direct real-time feedback about the course of the shift and creates motivation.

Shift meetings & handovers — Data basis for structured, fact-based conversations instead of gut feeling.

Structured improvement — In the long term, a database will be created that enables better planning and systematic KVP.

A reference customer summed up the difference in quality: Once the WLAN went down and no data was available, you immediately went back to your gut feeling — “Yes, it went okay.” This suddenly made it clear how much the quality of the conversations had changed.

3. Added value — in the short term, in the medium term, in terms of quality

In the short term: Low-hanging fruits via Pareto evaluations — 20% of the reasons for malfunctions cause 80% of downtime.

In the medium term: Better production planning based on real historical data instead of historical estimates.

Qualitatively: Away from gut feeling, towards figures, data, facts — a fundamentally different culture of conversation in production.

4. Data collection & connectivity

With a heterogeneous plant park, there are various options:

• Older machines: IoT gateway (additional hardware) collects signals and transmits them via WLAN or mobile communications to the European data center.
• Newer machines: Direct signal output from the controller via OPC UA or MQTT — without additional hardware.
• Machine controls: Depending on the manufacturer, an OPC interface can be retrofitted (example Heidenhain: approx. 790€). oee.ai has experience with a variety of control types.

The central signal is the quantity signal — this allows both availability (is the machine running?) as well as performance (is it running fast enough?) measure.

Disturbance index: Flexible, expandable at any time, cannot be manipulated on the shop floor. Environmental problems that the machine itself is not aware of (break times, staff shortages, tool changes) are entered manually via the tablet.

IT infrastructure recommendation: Production and office should run on separate networks. If one network fails, the other remains functional — especially in production, where downtime directly means loss of value creation.

5. An overview of the interface

The system is hierarchically structured: Individual machine → Division → Plant → Company. All levels show a configurable cockpit with widget-based real-time display.

Central elements: heartbeat line (target vs. actual in minute resolution), traffic light widget for the machine overview, Pareto charts on availability and performance levels, and flexible reporting (day, shift, weekly, order or product-related).

Access is exclusively via the browser — on a television, tablet, PC or smartwatch, without installation.

Data security: oee.ai works with DigitalOcean (European data center) and has all current security certifications. A penetration test with a customer in Asia was passed without any complaints.

‍