Soon, a new “motion insights and diagnostics” service from Swiss precision-drive specialist maxon called maxon MIND will help machine builders and plant engineers add condition monitoring and diagnostics capabilities to their equipment.
Electric motors are often the best sensor in a machine. After all, the motor signals a controller already reads reflect the condition of the entire drive system and its environment. maxon MIND turns those existing signals into diagnoses by monitoring the motor … and with it all the mechanical components connected downstream.
Developed in-house over years by an interdisciplinary team, maxon MIND works like this:
- There’s an application already driven by a maxon motor, gearbox, encoder, and controller(s).
- At the primary controller, a small software package installs and then runs a script to trigger a diagnostic cycle, generate references, record motor signals, and then enrich collected data with context and distillation for analysis.
- The system then sends the collected and condensed data over a secure internet connection (via MQTT, for example) to the Cloud-based maxon MIND platform. Once there, the data is analyzed by a machine-learning model and compared to the drive’s reference state. Deviations and anomalies are flagged.
- Integrated data, condition analyses, and historical trends are accessed through an interface made possible with an application programming interface (API) from maxon. So, personnel can pull results into plant dashboards and fleet-management systems. Diagnoses from the maxon MIND system are traceable, which is necessary in safety-critical design and regulated industries. Via the API, maxon MIND also integrates with the maxon Cloud.
Once maxon MIND is officially released to the industry (expected in 2027), machine builders aiming to leverage the service will first consult with maxon on their design load profiles, usage cycles, and environmental conditions. Then rapid prototyping and early testing (followed by system testing and validation under real operating conditions) will inform changes and the calibration of models. Final implementation will include optimization and fine-tuning of parameters or reference cycles.
Cybersecurity is ensured through the way collected machine data is anonymized and secured with maxon privacy measures complementing Cloud and security services. In short, maxon MIND doesn’t need to know who the end user is or what specific application is running. To uniquely identify the corresponding machine learning model, maxon MIND only needs data that’s easily anonymized — for example, a drive’s part and serial number.
Q&A with maxon MIND program lead
For more on how maxon MIND actually works, we asked business development manager and program lead for maxon MIND Claude Jaquemet a few questions. Here’s what he had to say.
Design World: Early literature detailing maxon MIND details how the software logs machine data during regular operation (and compares it to mechatronic system models) with a frequency that depends on the application and machine builder. But the maxon MIND system needs relatively small data volumes (not enormous datasets) thanks to the “integration of domain-specific expertise into the models” of how a motor, gearbox, and load physically relate.
What does domain-specific expertise mean in this context?
Jaquemet: In maxon MIND, domain-specific expertise means incorporating the known physical relationships and operating context of the mechatronic drive system into the model. This includes knowledge of the motor, controller, mechanical load, motion cycle, configuration, and relevant environmental conditions. It is interdisciplinary application knowledge, based on the long experience of maxon.
Design World: With this offering, a machine learning model is trained on the OEM’s own data about actual system behavior … plus, the models can be tuned so that a condition warning is genuinely a warning.
How does maxon MIND differentiate a real problem from slight and acceptable wear over time?
Jaquemet: With application-tailored models and high sensitivity. Therefore, maxon MIND doesn’t apply one universal wear threshold. It tracks the drive system relative to its application-specific initial state and visualizes the progression over time. Together with the machine builder, the model sensitivity and the boundary between acceptable condition and action-required condition are validated for the particular application. The result supports condition-based maintenance decisions.
Design World: We understand the architecture can be configured for different levels of complexity … and sensitivity can be adjusted for challenging situations involving cleanrooms, assemblies subject to vibrations, and extreme temperatures.
How exactly does maxon MIND distinguish problems from normal operating conditions?
Jaquemet: maxon MIND doesn’t rely on a universal definition of normal. A machine-learning model is created for the specific motor in its actual application and is trained on an initial reference state. Later measurements are compared with this application-specific model to detect deviations and wear trends. Also, various potential fault-modes are considered in the training of the initial reference state, to which the application is continuously compared. For example, to compensate for temperature effects, the applicable payload, the reference cycle itself, and the drivetrain configuration need to remain consistent during data acquisition.
Design World: Potential applications include those in laboratory automation (to protect sample integrity and throughput while holding quality consistent); medical devices (to prevent any situations capable of jeopardizing patient safety); and general industrial automation — for avoiding downtime, minimizing the waste of spare parts, and boosting overall equipment effectiveness (OEE).
In more general terms, what are the target applications?
Jaquemet: maxon MIND is designed for critical applications where an unexpected drive-system failure could affect personal safety, reliability, availability, process quality, or the user experience. Relevant application areas include medical devices, industrial and laboratory automation, and logistics. Technically, the application must allow repeatable diagnostic cycles under controlled and comparable load conditions. maxon MIND generally performs best where high sensitivity of the condition indication is required.
Design World: Can the motor signals reflect the condition of the whole powertrain, and how can the engineer localize a problem to one component?
Jaquemet: The electrical motor signals contain information influenced by the motor, gearbox, encoder, coupling, and driven mechanics. maxon MIND therefore assesses the condition of the complete drive system rather than only the motor. Component-level detection of faults is generally feasible, but it depends on the particular drivetrain. Beyond general component failure patterns, application-specific validation against known faults, inspections, or service findings help with the interpretation of the condition indication.
maxon | maxongroup.com
Leave a Reply
You must be logged in to post a comment.