The silent lever in mechanical engineering is called “condition-based maintenance” – not calendar-based maintenance.

Downtime is rarely just “bad luck”. There are usually warning signs beforehand.

In German SMEs, maintenance is often planned as follows: fixed intervals, fixed checklists, fixed spare parts. This looks organised. However, it is often expensive. Because real problems do not follow a calendar. They follow stress, wear and process deviations.

The question is not: “Have we carried out maintenance?”

The question is: “Did we do the right thing at the right time?”

Bosch Rexroth: Condition-based maintenance in its own factory

In a case study from its Homburg plant, Bosch Rexroth describes how condition-based maintenance was implemented on hydraulic test benches. The aim: fewer faults caused by contamination, fewer unnecessary maintenance measures, and more consistent quality. In one specific case, an efficiency increase of 5 % is cited.

This is not marketing “wishy-washy”, but precisely the lever that SMEs feel:

• fewer unplanned stoppages,

• fewer parts replaced ‘on a hunch’,

• greater process stability.

What constitutes “knowledge” here – and why AI makes sense in this context

Condition-based maintenance only works if you detect deviations early:

• Pressure profiles change,

• Temperature drifts,

• Vibrations increase,

• Cycle times become “erratic”.

AI excels at filtering such patterns from data. It detects deviations faster than the human eye. It recognises combinations that people overlook in everyday life. However: AI does not automatically tell you what you should do. That remains a human decision.

The SME toolkit: Getting started can be this simple

You don’t need a ‘major Industry 4.0 overhaul’. Start like this:

1. Choose a bottleneck system. The one that determines the line.

2. Measure 2–4 signals that you already have or can easily add.

3. Define 3 typical failure scenarios (e.g. leakage, contamination, bearings).

4. Determine: “What does 1 hour of downtime really cost us?”

5. Decide after 8–12 weeks: scale up or stop.

This is precisely where “AI-supported, but human-led” comes into play:

AI classifies signals. People assess causes. People set priorities.

What SMEs can take away from this

• Scheduled maintenance is often more expensive than necessary.

• A small pilot project can be enough to prove the benefits.

• AI helps with pattern recognition. People take the action.

• The goal is not ‘more data’. The goal is less downtime.