Gains:
- Knows which data indicates a malfunction
- Can make predictions with simple threshold monitoring
- Distinguishes what AI can and cannot do here
Most malfunctions do not occur suddenly; gives a signal beforehand. Increased vibration, temperature increase, change in current draw and change in sound character are the most common precursors.
Data to track
- Motor current: rising current may indicate blockage or bearing sticking.
- Bearing temperature: lack of lubrication or misalignment.
- Vibration: imbalance, wear, loosening.
- Aspiration pressure difference: filter clogging.
- Production efficiency: decreasing efficiency with the same setting is an indirect indicator of wear.
The role of artificial intelligence
Artificial intelligence follows the pattern in this data more patiently than a human: the detection of “the motor current of the C2 roller has been gradually increasing in the last three weeks” makes visible a trend that might have been overlooked.
However, artificial intelligence does not see the malfunction, it sees the data. If there is no sensor and no data is recorded, there is no prediction. The prerequisite for predictive maintenance is measurement; software is the second step.
The sentence "AI predicts failure" is true only if there is data. Measure first, model later. Building a model without measuring means making inferences from non-existent information.