maxon MIND
The motor as a sensor
The motor as a source for monitoring the condition of a machine: that is the goal of maxon MIND (Motion Insights and Diagnostics). The foundation is machine learning, physical expertise, and an open architecture.
Text Claude Jaquemet Photo Jeremias Wieland
In industrial drive technology, the motor has long been more than just an actuator. It provides valuable information about the condition of the entire system – pro- vided you know how to interpret the data correctly. This is exactly where maxon MIND comes in. The system uses motor signals to draw conclusions about the condition of the drive system and its en- vironment. This turns the motor into a sensor in the machine. maxon MIND analyzes data read from the controller during machine operation and detects patterns that indicate wear or malfunctions. It is not just about mon- itoring the motor itself, but the entire drive system and the associated mechan- ical components. This allows deviations to be detected early. The solution is particularly suitable for applications where potential failures must be detected early – for example, to increase personnel safety or minimize downtime. It also identifies problems
that remain hidden in conventional tests. In production lines with multiple ma- chines in sequence, maxon MIND detects even the smallest deviations before they affect downstream processes. The system also adds value in service operations when the actual condition of the equip- ment is critical for maintenance planning. Explainable, optimized AI Technologically, maxon MIND is based on explainable artificial intelligence. This makes diagnoses and predictions traceable – a key requirement for safety- critical and highly regulated applica- tions. The machine learning model de- veloped by maxon requires only small amounts of data and can run on simple hardware. This enables minimally in- vasive integration and reduces the ef- fort required for model training. As no computing infrastructure with graphics processing units (GPUs) is required,it can also offer significant cost advan-
tages. Existing applications can also be retrofitted. Data processing follows a clearly structured workflow. During machine operation, raw data is collected, prepro- cessed, and fed into the machine learning model. The model analyzes deviations from a healthy operating state. The re- sults are made available to customers via an Application Programming Interface (API). Glassbox instead of black box: every processing step remains traceable. High prediction quality A key feature of maxon MIND is the in- tegration of domain-specific expertise into the model. The physical relation- ships within a mechatronic system flow directly into the modeling process. This achieves high prediction quality with minimal data requirements. The system also learns from real- world applications. During commission- ing, an initial state is recorded that de-
36
maxon insights # 1
Made with FlippingBook Publishing Software