The manufacturing industry faces the challenge of operating increasingly complex production systems with maximum availability. Predictive maintenance has proven to be a key technology to proactively address wear and failures. Critical components can be calculated for replacement before larger unplanned interruptions or even damage to machine parts occur.
No prediction without precise measurement
To make informed predictions, one must first know the current state of a system as accurately as possible. What initially sounds trivial is often a greater challenge in engineering practice. Machines and systems are becoming increasingly complex, making direct, non-destructive measurement of critical parameters often either impossible or not feasible at reasonable costs.
This is where virtual sensor technology, also known as soft or inferential sensing, comes into play. This technique allows for inferring a desired characteristic based on easily measurable parameters, which cannot be directly measured with physical instruments or only with significant effort.
An interesting example from the industry is motor current analysis or Motor Current Signature Analytics (MCSA). The method is based on the fact that the current drawn by an AC motor is influenced not only by the applied voltage but also by the condition of the motor and the connected aggregates. Distortions in the current curve that cannot be explained by deformations in the voltage curve must therefore be caused by malfunctions in the system. The extent of the deviations can already indicate the severity of a disturbance. However, this alone is not enough – after all, engineers want to know exactly where the potential sources of error lie: Is there an imbalance or bearing damage, is cavitation occurring in pumps, or is there a stator or rotor fault? To conclude specific (mechanical) problems from the measured current and voltage curves, the detection of certain patterns in the deviations and linking them to specific sources of error is necessary.
Machine Learning and Predictions
When it comes to analyzing large amounts of data and detecting patterns, artificial intelligence can leverage its strengths through the subfield of machine learning. The technology supports maintenance teams in realizing predictive maintenance based on real-time data. This enables them to decide when to carry out which maintenance tasks and optimally coordinate these with production capacity.
With the combination of MCSA and AI-supported predictive maintenance, as used in Eaton's Motor Analytics, teams receive specific predictions about failure types for each individual motor – in a prioritized order and in the broader context of their electrical system. Subsequently, they can conduct deeper analyses with a pre-made dashboard to better understand performance deviations in the operation of machines and systems. Continuous digital data collection makes time-consuming manual surveys redundant. This not only relieves employees but also reduces the likelihood of errors.
By using MCSA solutions, companies can also predict both electrical and mechanical failures up to 30 percent earlier and 25 percent more accurately than with conventional condition monitoring systems. The AI-supported forecasting function provides directly usable insights without requiring extensive expertise in the respective system or data science experience.
Overall, it becomes clear that the combination of virtual sensor technology, motor current analysis, and AI-supported predictive maintenance gives companies a significant advantage in the availability and efficiency of their systems. Early, precise insights into the condition of critical components enable not only planned maintenance but also reduce failure risks and operating costs. Thus, digital, data-driven maintenance becomes a central success factor for the manufacturing industry of the future.
Contact:
